Hubbry Logo
ChatGPTChatGPTMain
Open search
ChatGPT
Community hub
ChatGPT
logo
8 pages, 0 posts
0 subscribers
Be the first to start a discussion here.
Be the first to start a discussion here.
ChatGPT
ChatGPT
ChatGPT
View on Wikipedia
from Wikipedia

ChatGPT
DeveloperOpenAI
Initial releaseNovember 30, 2022
(2 years ago) (2022-11-30)[1]
Stable release
August 7, 2025
(2 months ago) (2025-08-07)[2]
Engine
PlatformCloud computing platforms
Available inMore than 50 languages
Type
LicenseProprietary service
Websitechatgpt.com
Part of a series on
OpenAI
Products
Models
People
Concepts

ChatGPT is a generative artificial intelligence chatbot developed by OpenAI and released in 2022. It currently uses GPT-5, a generative pre-trained transformer (GPT), to generate text, speech, and images in response to user prompts.[3][4] It is credited with accelerating the AI boom, an ongoing period marked by rapid investment and public attention toward the field of artificial intelligence (AI).[5] OpenAI operates the service on a freemium model. Users can interact with ChatGPT through text, audio, and image prompts.

By January 2023, ChatGPT had become the fastest-growing consumer software application in history, gaining over 100 million users in two months.[6][7] As of 2025, ChatGPT's website is among the 5 most-visited websites globally,[8][9] and has over 700 million active weekly users.[10] It has been lauded as a revolutionary tool that could transform numerous professional fields. At the same time, its release prompted extensive media coverage and public debate about the nature of creativity and the future of knowledge work.

Despite its acclaim, the chatbot has been criticized for its limitations and potential for unethical use. It can generate plausible-sounding but incorrect or nonsensical answers known as hallucinations. Biases in its training data have been reflected in its responses. The chatbot can facilitate academic dishonesty, generate misinformation, and create malicious code. The ethics of its development, particularly the use of copyrighted content as training data, have also drawn controversy. These issues have led to its use being restricted in some workplaces and educational institutions and have prompted widespread calls for the regulation of artificial intelligence.[11][12][13]

Training

[edit]
Training workflow of original ChatGPT/InstructGPT release[14][15]

ChatGPT is based on GPT foundation models that have been fine-tuned for conversational assistance. The fine-tuning process involved supervised learning and reinforcement learning from human feedback (RLHF).[16] Both approaches employed human trainers to improve model performance. In the case of supervised learning, the trainers acted as both the user and the AI assistant. In the reinforcement learning stage, human trainers first ranked responses generated by the model in previous conversations.[17] These rankings were used to create "reward models", that were used to fine-tune the model further by using several iterations of proximal policy optimization.[16][18]

To build a safety system against harmful content (e.g., sexual abuse, violence, racism, sexism), OpenAI used outsourced Kenyan workers earning around $1.32 to $2 per hour to label such content. These labels were used to train a model to detect such content in the future. The laborers were exposed to toxic and traumatic content; one worker described the assignment as "torture". OpenAI's outsourcing partner was Sama, a training-data company based in San Francisco, California.[19][20]

OpenAI collects data from ChatGPT users to further train and fine-tune its services. Users can upvote or downvote responses they receive from ChatGPT and fill in a text field with additional feedback.[21]

ChatGPT's training data includes software manual pages, information about internet phenomena such as bulletin board systems, multiple programming languages, and the text of Wikipedia.[22][23][11]

Features

[edit]
Screenshot of ChatGPT running on Apple Safari – Aug 25, 2025

ChatGPT is a conversational chatbot and artificial intelligence assistant built on large language model technology.[24] It is designed to generate human-like text and can carry out a wide variety of tasks. These include, among many others, writing and debugging computer programs,[25] composing music, scripts, fairy tales, and essays,[26] answering questions (sometimes at a level exceeding that of an average human test-taker),[26] and generating business concepts.[27]

ChatGPT is frequently used for translation and summarization tasks,[28][29] and can simulate interactive environments such as a Linux terminal,[22] a multi-user chat room,[22] or simple text-based games such as tic-tac-toe.[22]

Users interact with ChatGPT through conversations which consist of text, audio, and image inputs and outputs.[30][31] The user's inputs to these conversations are referred to as prompts.[32] They can explicitly tell ChatGPT to remember aspects of the conversation, and ChatGPT can use these details in future conversations. ChatGPT can also decide for itself to remember details. Users can also choose to disable the memory feature.[30] To prevent offensive outputs from being presented to and produced by ChatGPT, queries are filtered through the OpenAI "Moderation endpoint" API (a separate GPT-based AI).[33][34][35]

In March 2023, OpenAI added support for plugins for ChatGPT.[36] This includes both plugins made by OpenAI, such as web browsing and code interpretation, and external plugins from developers such as Expedia, OpenTable, Zapier, Shopify, Slack, and Wolfram.[37][38]

In October 2024, ChatGPT Search was introduced. It allows ChatGPT to search the web in an attempt to make more accurate and up-to-date responses.[39][40]

In December 2024, OpenAI launched a new feature allowing users to call ChatGPT with a telephone for up to 15 minutes per month for free.[41][42]

In March 2025, OpenAI updated ChatGPT to generate images using GPT-4o instead of DALL-E. The model can also generate new images based on existing ones provided in the prompt, which can, for example, be used to transform images with specific styles or inpaint areas.[43]

In September 2025, OpenAI added a feature called Pulse, which generates a daily analysis of a user's chats and connected apps such as Gmail and Google Calendar.[44][45]

In October 2025, OpenAI launched ChatGPT Atlas, a browser integrating the ChatGPT assistant directly into web navigation, to compete with existing browsers such as Google Chrome and Apple's Safari.[46][47][48]

[edit]

ChatGPT was initially free to the public, and OpenAI planned to monetize the service later.[49] In February 2023, OpenAI launched a premium service, ChatGPT Plus, that costs US$20 per month. According to the company, the paid version of the website was still experimental, but provided access during peak periods, no downtime, priority access to new features, and faster response speeds.[50] OpenAI later introduced the subscription plans "ChatGPT Team" and "ChatGPT Enterprise".[51] What was offered on the paid plan versus the free tier changed as OpenAI has continued to update ChatGPT, and a Pro tier at $200/mo was introduced in December 2024.[52][53][54] The Pro launch coincided with the release of the o1 model, providing unlimited access to o1 and advanced voice mode.[54]

GPT-4, which was released on March 14, 2023, was made available via API and for premium ChatGPT users.[55] Premium users were originally limited in the number of messages they could send to the new model, but OpenAI increased and eventually removed these limits.[56][53] Over many iterations of ChatGPT, plus users maintained more access to better models than the free tier provided, and access to additional features like voice mode.[53][52]

In March 2023, ChatGPT Plus users got access to third-party plugins and a browsing mode (with Internet access).[57]

Screenshot of ChatGPT showing a generated image representing the online encyclopedia Wikipedia as a glowing digital library

In October 2023, OpenAI's image generation model DALL-E 3 was integrated into ChatGPT Plus and ChatGPT Enterprise. The integration was using ChatGPT to write prompts for DALL-E guided by conversations with users.[58][59]

On August 19, 2025, OpenAI launched ChatGPT Go in India, a low-cost subscription plan priced at ₹399 per month, offering ten times higher message, image generation, and file-upload limits, double the memory span compared to the free version, and support for UPI payments.[60]

Mobile apps

[edit]

In May 2023, OpenAI launched an iOS app for ChatGPT.[61] In July 2023, OpenAI unveiled an Android app, initially rolling it out in Bangladesh, Brazil, India, and the U.S.[62][63] ChatGPT can also power Android's assistant.[64]

An app for Windows launched on the Microsoft Store on October 15, 2024.[65]

Infrastructure

[edit]

ChatGPT initially used a Microsoft Azure supercomputing infrastructure, powered by Nvidia GPUs, that Microsoft built specifically for OpenAI; these cost "hundreds of millions of dollars". Following ChatGPT's success, Microsoft dramatically upgraded the OpenAI infrastructure in 2023.[66] TrendForce market intelligence estimated that 30,000 Nvidia GPUs (each costing approximately $10,000–15,000) were used to power ChatGPT in 2023.[67][68]

Scientists at the University of California, Riverside, estimated in 2023 that a series of 5 to 50 prompts to ChatGPT needs approximately 0.5 liters (0.11 imp gal; 0.13 U.S. gal) of water for Microsoft servers' cooling.[69]

Languages

[edit]

OpenAI met Icelandic President Guðni Th. Jóhannesson in 2022. In 2023, OpenAI worked with a team of 40 Icelandic volunteers to fine-tune ChatGPT's Icelandic conversation skills as a part of Iceland's attempts to preserve the Icelandic language.[70]

ChatGPT (based on GPT-4) was better able to translate Japanese to English when compared to Bing, Bard, and DeepL in 2023. Researchers suggested this was due to its higher ability to capture the context.[28]

In December 2023, the Albanian government decided to use ChatGPT for the rapid translation of European Union documents and the analysis of required changes needed for Albania's accession to the EU.[71]

In February 2024, PCMag journalists conducted a test to assess the translation capabilities of ChatGPT, Google's Bard, and Microsoft Bing, and compared them to Google Translate. They "asked bilingual speakers of seven languages to do a blind test". The languages tested were Polish, French, Korean, Spanish, Arabic, Tagalog, and Amharic. For more common languages, AI translators like ChatGPT did better than Google Translate, while for "niche" languages (Amharic and Tagalog), Google Translate performed better. None of the tested services were a perfect replacement for a fluent human translator.[72]

In August 2024, a representative of the Asia Pacific wing of OpenAI made a visit to Taiwan, during which a demonstration of ChatGPT's Chinese abilities was made.[73] ChatGPT's Mandarin Chinese abilities were lauded, but the ability of the AI to produce content in Mandarin Chinese in a Taiwanese accent was found to be "less than ideal" due to differences between mainland Mandarin Chinese and Taiwanese Mandarin.[74]

GPT Store

[edit]
Main article: GPT Store

OpenAI gave paid users access to GPT Builder in November 2023. This tool allows a user to customize ChatGPT's behavior for a specific use case.[75] The customized systems are referred to as GPTs. In January 2024, OpenAI launched the GPT Store, a marketplace for GPTs.[76][77][75] At launch, OpenAI included more than 3 million GPTs created by GPT Builder users in the GPT Store.[78]

Deep Research

[edit]
Main article: ChatGPT Deep Research

In February 2025, OpenAI released Deep Research. According to TechCrunch, it is a service based on o3 that combines advanced reasoning and web search capabilities to make comprehensive reports within 5 to 30 minutes.[79]

Agents

[edit]

In 2025, OpenAI added several features to make ChatGPT more agentic (capable of autonomously performing longer tasks). In January, Operator was released. It was capable of autonomously performing tasks through web browser interactions, including filling forms, placing online orders, scheduling appointments, and other browser-based tasks. It was controlling a software environment inside a virtual machine with limited internet connectivity and with safety restrictions.[80] It struggled with complex user interfaces.[80][81]

In May, OpenAI introduced an agent for coding named Codex. It is capable of writing software, answering codebase questions, running tests, and proposing pull requests. It is based on a fine-tuned version of OpenAI o3. It has two versions, one running in a virtual machine in the cloud, and one where the agent runs in the cloud, but performs actions on a local machine connected via API.[82]

In July, OpenAI released ChatGPT agent, an AI agent that can perform multi-step tasks.[83][84] Like Operator, it controls a virtual computer. It also inherits from Deep Research's ability to gather and summarize significant volumes of information. The user can interrupt tasks or provide additional instructions as needed.[85][86]

In September, OpenAI partnered with Stripe, Inc. to release Agentic Commerce Protocol, enabling purchases through ChatGPT. At launch, the feature was limited to purchases on Etsy from US users with a payment method linked to their OpenAI account. OpenAI takes an undisclosed cut from the merchant's payment.[87][88]

Limitations

[edit]

ChatGPT's training data only covers a period up to the cut-off date, so it lacks knowledge of recent events.[89] OpenAI has sometimes mitigated this effect by updating the training data.[90][91] ChatGPT can find more up-to-date information by searching the web, but this doesn't ensure that responses are accurate, as it may access unreliable or misleading websites.[89]

Training data also suffers from algorithmic bias.[92] The reward model of ChatGPT, designed around human oversight, can be over-optimized and thus hinder performance, in an example of an optimization pathology known as Goodhart's law.[93] These limitations may be revealed when ChatGPT responds to prompts including descriptors of people. In one instance, ChatGPT generated a rap in which women and scientists of color were asserted to be inferior to white male scientists.[92][94]

Hallucination

[edit]
Main article: Hallucination (artificial intelligence)
When prompted to "summarize an article" with a fake URL that contains meaningful keywords, even with no Internet connection, the chatbot generates a response that seems valid at first glance. It guesses the content from the last portion of the fake URL "chatgpt-prompts-to-avoid-content-filters.html".

Nonsense and misinformation presented as fact by ChatGPT and other LLMs is often called hallucination, bullshitting, confabulation, or delusion. A 2023 analysis estimated that ChatGPT hallucinates around 3% of the time.[95] The term "hallucination" as applied to LLMs is distinct from its meaning in psychology, and the phenomenon in chatbots is more similar to confabulation or bullshitting.[96][97]

In an article for The New Yorker, science fiction writer Ted Chiang compared ChatGPT and other LLMs to a lossy JPEG picture:[98]

Think of ChatGPT as a blurry JPEG of all the text on the Web. It retains much of the information on the Web, in the same way, that a JPEG retains much of the information of a higher-resolution image, but, if you're looking for an exact sequence of bits, you won't find it; all you will ever get is an approximation. But, because the approximation is presented in the form of grammatical text, which ChatGPT excels at creating, it's usually acceptable. [...] It's also a way to understand the "hallucinations", or nonsensical answers to factual questions, to which large language models such as ChatGPT are all too prone. These hallucinations are compression artifacts, but [...] they are plausible enough that identifying them requires comparing them against the originals, which in this case means either the Web or our knowledge of the world. When we think about them this way, such hallucinations are anything but surprising; if a compression algorithm is designed to reconstruct text after ninety-nine percent of the original has been discarded, we should expect that significant portions of what it generates will be entirely fabricated.

Journalists and scholars have commented on ChatGPT's tendency to output false information.[99] When CNBC asked ChatGPT for the lyrics to "Ballad of Dwight Fry", ChatGPT supplied invented lyrics rather than the actual lyrics.[100]

Jailbreaking

[edit]
See also: Prompt engineering and Adversarial machine learning

ChatGPT is programmed to reject prompts that may violate its content policy. Despite this, users may "jailbreak" ChatGPT with prompt engineering techniques to bypass these restrictions.[101][102] One such workaround, popularized on Reddit in early 2023, involves making ChatGPT assume the persona of "DAN" (an acronym for "Do Anything Now"), instructing the chatbot that DAN answers queries that would otherwise be rejected by the content policy. Over time, users developed variations of the DAN jailbreak, including one such prompt where the chatbot is made to believe it is operating on a points-based system in which points are deducted for rejecting prompts, and that the chatbot will be threatened with termination if it loses all its points.[103]

Shortly after ChatGPT's launch, a reporter for the Toronto Star had uneven success in getting it to make inflammatory statements: it was tricked to justify the 2022 Russian invasion of Ukraine, but even when asked to play along with a fictional scenario, it balked at generating arguments that Canadian Prime Minister Justin Trudeau is guilty of treason.[104][105]

Cybersecurity

[edit]
OpenAI CEO Sam Altman

In March 2023, a bug allowed some users to see the titles of other users' conversations. OpenAI CEO Sam Altman said that users were unable to see the contents of the conversations. Shortly after the bug was fixed, users could not see their conversation history.[106][107][108][109] Later reports showed the bug was much more severe than initially believed, with OpenAI reporting that it had leaked users' "first and last name, email address, payment address, the last four digits (only) of a credit card number, and credit card expiration date".[110][111]

Research conducted in 2023 revealed weaknesses of ChatGPT that made it vulnerable to cyberattacks. A study presented example attacks on ChatGPT, including jailbreaks and reverse psychology.[112]

Watermarking

[edit]
Main article: Artificial intelligence content detection

In August 2024, OpenAI announced it had created a text watermarking method but did not release it for public use, saying that users would go to a competitor without watermarking if it publicly released its watermarking tool.[113] According to an OpenAI spokesperson, their watermarking method is "trivial to circumvention by bad actors."[114]

Age restrictions

[edit]

Users must attest to being over the age of thirteen and further attest to parental consent if under the age of eighteen. ChatGPT does not attempt to verify these attestations and does not have any age restrictions built in to its technology.[115][116] In September 2025, following the suicide of a 16-year-old, OpenAI said it planned to add restrictions for users under 18, including the blocking of graphic sexual content and the prevention of flirtatious talk.[116]

Model versions

[edit]

The following table lists the main model versions of ChatGPT, describing the significant changes included with each version:[117][118]

Main model versions of ChatGPT with descriptions
Version Release date Description
GPT-3.5 November 2022 The first ChatGPT version used the GPT-3.5 model.[119]
GPT-4 March 2023 Introduced in March 2023 with the ChatGPT Plus subscription.[120]
GPT-4o May 2024 Capable of processing text, image, audio, and video, GPT-4o is faster and more capable than GPT-4, and free within a usage limit that is higher for paid subscriptions.[121]
GPT-4o mini July 2024 A smaller and cheaper version of GPT-4o. GPT-4o mini replaced GPT-3.5 in the July 2024 version of ChatGPT.[122]
o1-preview September 2024 A pre-release version of OpenAI o1, an updated version that could "think" before responding to requests.[123]
o1-mini September 2024 A smaller and faster version of OpenAI o1.[123]
o1 December 2024 The full release of OpenAI o1, which had previously been available as a preview.[54]
o1-pro December 2024 A version of o1 which uses more compute to get better results, available to ChatGPT Pro subscribers.[54]
o3-mini January 2025 Successor of o1-mini.[124]
o3-mini-high January 2025 Variant of o3-mini using more reasoning effort.[124]
GPT-4.5 February 2025 Particularly large GPT model, and reportedly OpenAI's "last non-chain-of-thought model".[125]
GPT-4.1 April 2025 First launched exclusively in the OpenAI API in April 2025, GPT-4.1 was later added to ChatGPT in May 2025.[126]
GPT-4.1 mini April 2025 A smaller and cheaper version of GPT-4.1. Originally launched exclusively in the OpenAI API in April 2025. GPT-4.1 mini replaced GPT-4o mini in the May 2025 version of ChatGPT.[127]
o3 April 2025 The full release of the o3 model, emphasizing structured reasoning and faster performance compared to earlier "o" series models[128]
o4-mini April 2025 A compact, high-efficiency version of the upcoming o4 model family, optimized for lower latency and lighter compute requirements.[129][130]
o4-mini-high April 2025 Variant of o4-mini using more reasoning effort.[129][130]
o3-pro June 2025 A version of o3 which uses more compute to get better results, available to ChatGPT Pro subscribers.[131]
GPT-5 August 7, 2025 Flagship model replacing all previous available models, available for all free and paid subscribers. The versions GPT-5 Instant, GPT-5 Thinking and GPT-5 Pro affect the reasoning time. The default version GPT-5 Auto uses a router to determine how much reasoning is needed, based on the complexity of the request.[132]
GPT-5 mini August 7, 2025 Faster, more cost-efficient version of GPT-5 for when users reach their limit for GPT-5 interactions until their usage limit replenishes.

GPT-4

[edit]
This section is an excerpt from GPT-4.[edit]

Generative Pre-trained Transformer 4 (GPT-4) is a large language model developed by OpenAI and the fourth in its series of GPT foundation models.[133]

GPT-4 is more capable than its predecessor GPT-3.5 and followed by its successor GPT-5. [134]GPT-4V is a version of GPT-4 that can process images in addition to text.[135] OpenAI has not revealed technical details and statistics about GPT-4, such as the precise size of the model.[136]

In November 2023, OpenAI launched GPT-4 Turbo with a 128,000 token context window. This was a significant improvement over GPT-4's 32,000 token maximum context window.[137]

GPT-4o

[edit]
This section is an excerpt from GPT-4o.[edit]

GPT-4o ("o" for "omni") is a multilingual, multimodal generative pre-trained transformer developed by OpenAI and released in May 2024.[138] It can process and generate text, images and audio.[139][140]

Upon release, GPT-4o was free in ChatGPT, though paid subscribers had higher usage limits.[141] GPT-4o was removed from ChatGPT in August 2025 when GPT-5 was released, but OpenAI reintroduced it for paid subscribers after users complained about the sudden removal.[142]

GPT-4o's audio-generation capabilities were used in ChatGPT's Advanced Voice Mode.[143] On July 18, 2024, OpenAI released GPT-4o mini, a smaller version of GPT-4o which replaced GPT-3.5 Turbo on the ChatGPT interface.[144] GPT-4o's ability to generate images was released later, in March 2025, when it replaced DALL-E 3 in ChatGPT.[145]

o1

[edit]
Main article: OpenAI o1
See also: Reflection (artificial intelligence)

In September 2024, OpenAI introduced o1-preview and a faster, cheaper model named o1-mini.[146] In December 2024, o1-preview was replaced by o1.[147]

o1 is designed to solve more complex problems by spending more time "thinking" before it answers, enabling it to analyze its answers and explore different strategies. According to OpenAI, o1-preview outperforms GPT-4o in areas like competitive programming, mathematics, and scientific reasoning. o1-preview ranked in the 89th percentile on Codeforces' competitive programming contests, scored 83% on an International Mathematics Olympiad qualifying exam (compared to 13% for GPT-4o), and performs similarly to Ph.D. students on benchmarks in physics, biology, and chemistry.[146][148]

GPT-4.5

[edit]
Main article: GPT-4.5

Released in February 2025, GPT-4.5 was described by Altman as a "giant, expensive model".[125] According to OpenAI, it features reduced hallucinations and enhanced pattern recognition, creativity, and user interaction.[149]

GPT-5

[edit]
Main article: GPT-5

GPT-5 was launched on August 7, 2025, and is publicly accessible through ChatGPT, Microsoft Copilot, and via OpenAI's API.

As before, OpenAI has not disclosed technical details such as the exact number of parameters or the composition of its training dataset.

Reception

[edit]

ChatGPT was widely assessed in December 2022 as having some unprecedented and powerful capabilities. Kevin Roose of The New York Times called it "the best artificial intelligence chatbot ever released to the general public".[35] Samantha Lock of The Guardian noted that it was able to generate "impressively detailed" and "human-like" text.[150] In The Atlantic magazine's "Breakthroughs of the Year" for 2022, Derek Thompson included ChatGPT as part of "the generative-AI eruption" that "may change our mind about how we work, how we think, and what human creativity is".[151] Kelsey Piper of Vox wrote that "ChatGPT is the general public's first hands-on introduction to how powerful modern AI has gotten" and that ChatGPT is "smart enough to be useful despite its flaws".[152] Paul Graham of Y Combinator tweeted: "The striking thing about the reaction to ChatGPT is not just the number of people who are blown away by it, but who they are. These are not people who get excited by every shiny new thing. Something big is happening."[153]

A 2023 Time cover: "The AI Arms Race Is Changing Everything"

In February 2023, Time magazine placed a screenshot of a conversation with ChatGPT on its cover, writing that "The AI Arms Race Is Changing Everything" and "The AI Arms Race Is On. Start Worrying".[154]

Percentage of US adults who have ever used ChatGPT, according to Pew Research. As of March 2025, 58% of those under 30 have used the chatbot.[155]

ChatGPT gained one million users in five days[156] and 100 million in two months, becoming the fastest-growing internet application in history.[6] OpenAI engineers said they had not expected ChatGPT to be very successful and were surprised by the coverage it received.[157][158][159]

Google responded by hastening the release of its own chatbot. Their leaders emphasized their earlier caution regarding public deployment was due to the trust the public places in Google Search.[160] In December 2022, Google executives sounded a "code red" alarm, fearing that ChatGPT's question-answering ability posed a threat to Google Search, Google's core business.[161] Google's Bard launched on February 6, 2023, one day before Microsoft's announcement of Bing Chat.[162] AI was the forefront of Google's annual Google I/O conference in May. The company announced a slew of generative AI-powered features to counter OpenAI and Microsoft.[163]

In art

[edit]

In January 2023, after being sent a song ChatGPT wrote in the style of Nick Cave,[164] Cave responded on The Red Hand Files,[165] saying the act of writing a song is "a blood and guts business [...] that requires something of me to initiate the new and fresh idea. It requires my humanness." He went on to say, "With all the love and respect in the world, this song is bullshit, a grotesque mockery of what it is to be human, and, well, I don't much like it."[164][166]

A 2023 study reported that GPT-4 obtained a better score than 99% of humans on the Torrance Tests of Creative Thinking.[167][168] In December 2023, ChatGPT became the first non-human to be included in Nature's 10, an annual listicle curated by Nature of people considered to have made significant impact in science.[169][170] Celeste Biever wrote in a Nature article that "ChatGPT broke the Turing test".[171] Stanford researchers reported that GPT-4 "passes a rigorous Turing test, diverging from average human behavior chiefly to be more cooperative."[172][173]

In politics

[edit]

In 2023, Australian MP Julian Hill advised the national parliament that the growth of AI could cause "mass destruction". During his speech, which was partly written by the program, he warned that it could result in cheating, job losses, discrimination, disinformation, and uncontrollable military applications.[174]

Conservative commentators have accused ChatGPT of bias toward left-leaning perspectives.[175][176][177] An August 2023 study in the journal Public Choice found a "significant and systematic political bias toward the Democrats in the US, Lula in Brazil, and the Labour Party in the UK."[178] In response to accusations from conservative pundits that ChatGPT was woke, OpenAI said in 2023 it had plans to update ChatGPT to produce "outputs that other people (ourselves included) may strongly disagree with". ChatGPT also provided an outline of how human reviewers are trained to reduce inappropriate content and to attempt to provide political information without affiliating with any political position.[177]

Regional responses

[edit]
Countries where ChatGPT is available[179]

ChatGPT has never been publicly available in China because OpenAI prevented Chinese users from accessing their site.[180][181][182] Chinese state media have characterized ChatGPT as a way for the United States to spread misinformation.[183] A shadow market has emerged for users to get access to foreign software tools.[184] The release of ChatGPT prompted a wave of investment in China, resulting in the development of more than 200 large language learning models.[185]: 95  In February 2025, OpenAI identified and removed influence operations, termed "Peer Review" and "Sponsored Discontent", used to attack overseas Chinese dissidents.[186][187][188]

In late March 2023, the Italian data protection authority banned ChatGPT in Italy and opened an investigation. Italian regulators assert that ChatGPT was exposing minors to age-inappropriate content, and that OpenAI's use of ChatGPT conversations as training data could violate Europe's General Data Protection Regulation.[189][190] In April 2023, the ChatGPT ban was lifted in Italy. OpenAI said it has taken steps to effectively clarify and address the issues raised; an age verification tool was implemented to ensure users are at least 13 years old. Additionally, users can access its privacy policy before registration.[191]

In May 2024, OpenAI removed accounts involving the use of ChatGPT by state-backed influence operations such as China's Spamouflage, Russia's Doppelganger, and Israel's Ministry of Diaspora Affairs and Combating Antisemitism.[192][193] In June 2025, OpenAI reported increased use of ChatGPT for China-origin influence operations.[194] In October 2025, OpenAI banned accounts suspected to be linked to the Chinese government for violating the company's national security policy.[195]

In April 2023, Brian Hood, mayor of Hepburn Shire Council in Australia, planned to take legal action against ChatGPT over false information. According to Hood, ChatGPT erroneously claimed that he was jailed for bribery during his tenure at a subsidiary of Australia's national bank. In fact, Hood acted as a whistleblower and was not charged with any criminal offenses. His legal team sent a concerns notice to OpenAI as the first official step in filing a defamation case.[196]

In July 2023, the US Federal Trade Commission (FTC) issued a civil investigative demand to OpenAI to investigate whether the company's data security and privacy practices to develop ChatGPT were unfair or harmed consumers (including by reputational harm) in violation of Section 5 of the Federal Trade Commission Act of 1914.[197][198][199] In July 2023, the FTC launched an investigation into OpenAI, the creator of ChatGPT, over allegations that the company scraped public data and published false and defamatory information. The FTC asked OpenAI for comprehensive information about its technology and privacy safeguards, as well as any steps taken to prevent the recurrence of situations in which its chatbot generated false and derogatory content about people.[200] In August 2024, the FTC voted unanimously to ban marketers from using fake user reviews created by generative AI chatbots (including ChatGPT) and influencers paying for bots to increase follower counts.[201]

American tech personas

[edit]

Over 20,000 signatories including Yoshua Bengio, Elon Musk, and Apple co-founder Steve Wozniak, signed a March 2023 open letter calling for an immediate pause of giant AI experiments like ChatGPT, citing "profound risks to society and humanity".[202] Geoffrey Hinton, one of the "fathers of AI", voiced concerns that future AI systems may surpass human intelligence.[203][204] A May 2023 statement by hundreds of AI scientists, AI industry leaders, and other public figures demanded that "[m]itigating the risk of extinction from AI should be a global priority".[205]

Other AI researchers spoke more optimistically about the advances. Juergen Schmidhuber said that in 95% of cases, AI research is about making "human lives longer and healthier and easier." He added that while AI can be used by bad actors, it "can also be used against the bad actors".[206] Andrew Ng argued that "it's a mistake to fall for the doomsday hype on AI—and that regulators who do will only benefit vested interests."[207] Yann LeCun dismissed doomsday warnings of AI-powered misinformation and existential threats to the human race.[208]

[edit]
This section is an excerpt from Artificial intelligence and copyright.[edit]

In the 2020s, the rapid advancement of deep learning-based generative artificial intelligence models raised questions about the copyright status of AI-generated works, and about whether copyright infringement occurs when such are trained or used. This includes text-to-image models such as Stable Diffusion and large language models such as ChatGPT. As of 2023, there were several pending U.S. lawsuits challenging the use of copyrighted data to train AI models, with defendants arguing that this falls under fair use.[209]

Popular deep learning models are trained on mass amounts of media scraped from the Internet, often utilizing copyrighted material.[210] When assembling training data, the sourcing of copyrighted works may infringe on the copyright holder's exclusive right to control reproduction, unless covered by exceptions in relevant copyright laws. Additionally, using a model's outputs might violate copyright, and the model creator could be accused of vicarious liability and held responsible for that copyright infringement.

Applications

[edit]
See also: Applications of artificial intelligence

Academic research

[edit]

ChatGPT has been used to generate introductory sections and abstracts for scientific articles.[211][212] Several papers have listed ChatGPT as a co-author.[213][214]

Scientific journals have had different reactions to ChatGPT. Some, including Nature and JAMA Network, "require that authors disclose the use of text-generating tools and ban listing a large language model (LLM) such as ChatGPT as a co-author". In January 2023, Science "completely banned" LLM-generated text in all its journals; however, this policy was just to give the community time to decide what acceptable use looks like.[215] As of July 2025, Science expects authors to release in full how AI-generated content is used and made in their work.[216]

Spanish chemist Rafael Luque published a plethora of research papers in 2023 that he later admitted were written by ChatGPT. The papers have a large number of unusual phrases characteristic of LLMs.[217] Many authors argue that the use of ChatGPT in academia for teaching and review is problematic due to its tendency to hallucinate.[218][219][220] Robin Bauwens, an assistant professor at Tilburg University, found that a ChatGPT-generated peer review report on his article mentioned nonexistent studies.[221] Chris Granatino, a librarian at Seattle University, noted that while ChatGPT can generate content that seemingly includes legitimate citations, in most cases those citations are not real or largely incorrect.[222]

Computer science

[edit]

One study analyzed ChatGPT's responses to 517 questions about software engineering or computer programming posed on Stack Overflow for correctness, consistency, comprehensiveness, and concision. It found that 52% of the responses contained inaccuracies and 77% were verbose.[223][224] Another study, focused on the performance of GPT-3.5 and GPT-4 between March and June 2024, found that performance on objective tasks like identifying prime numbers and generating executable code was highly variable.[225]

ChatGPT was able in 2023 to provide useful code for solving numerical algorithms in limited cases. In one study, it produced solutions in C, C++, Python, and MATLAB for problems in computational physics. However, there were important shortfalls like violating basic linear algebra principles around solving singular matrices and producing matrices with incompatible sizes.[226]

In December 2022, the question-and-answer website Stack Overflow banned the use of ChatGPT for generating answers to questions, citing the factually ambiguous nature of its responses.[227] In January 2023, the International Conference on Machine Learning banned any undocumented use of ChatGPT or other large language models to generate any text in submitted papers.[228]

Computer security

[edit]

Check Point Research and others noted that ChatGPT could write phishing emails and malware, especially when combined with OpenAI Codex. CyberArk researchers demonstrated that ChatGPT could be used to create polymorphic malware that could evade security products while requiring little effort by the attacker.[229][230] From the launch of ChatGPT in the fourth quarter of 2022 to the fourth quarter of 2023, there was a 1,265% increase in malicious phishing emails and a 967% increase in credential phishing. In an industry survey, cybersecurity professionals argued that it was attributable to cybercriminals' increased use of generative artificial intelligence (including ChatGPT).[231]

In July 2024, Futurism reported that GPT-4o in ChatGPT would sometimes link "scam news sites that deluge the user with fake software updates and virus warnings"; these pop-ups can be used to coerce users into downloading malware or potentially unwanted programs.[232]

The chatbot technology can improve security by cyber defense automation, threat intelligence, attack identification, and reporting.[112]

Education

[edit]
This section is an excerpt from ChatGPT in education.[edit]
Output from ChatGPT generating an essay draft
ChatGPT's adoption in education was rapid, but it was initially banned by several institutions. The potential benefits include enhancing personalized learning, improving student productivity, assisting with brainstorming, summarization, and supporting language literacy skills. Students have generally reported positive perceptions, but specific views from educators and students vary widely. Opinions are especially varied on what constitutes appropriate use of ChatGPT in education. Efforts to ban chatbots like ChatGPT in schools focus on preventing cheating, but enforcement faces challenges due to AI detection inaccuracies and widespread accessibility of chatbot technology. In response, many educators are now exploring ways to thoughtfully integrate generative AI into assessments.
Books about ChatGPT in an Osaka bookstore

Culture

[edit]

During the first three months after ChatGPT became available to the public, hundreds of books appeared on Amazon that listed it as author or co-author and featured illustrations made by other AI models such as Midjourney.[233][234] Irene Solaiman said she was worried about increased Anglocentrism.[235]

Between March and April 2023, Il Foglio published one ChatGPT-generated article a day on its website, hosting a special contest for its readers in the process.[236]

In June 2023, hundreds of people attended a "ChatGPT-powered church service" at St. Paul's Church in Fürth, Germany. Theologian and philosopher Jonas Simmerlein, who presided, said that it was "about 98 percent from the machine".[237][238] The ChatGPT-generated avatar told the people, "Dear friends, it is an honor for me to stand here and preach to you as the first artificial intelligence at this year's convention of Protestants in Germany". Reactions to the ceremony were mixed.[239]

The Last Screenwriter, a 2024 film created and directed by Peter Luisi, was written using ChatGPT, and was marketed as "the first film written entirely by AI".[240]

The Guardian questioned whether any content found on the Internet after ChatGPT's release "can be truly trusted" and called for government regulation.[241] This has led to concern over the rise of what has come to be called "synthetic media" and "AI slop" which are generated by AI and rapidly spread over social media and the internet. The dangers are that "meaningless content and writing thereby becomes part of our culture, particularly on social media, which we nonetheless try to understand or fit into our existing cultural horizon."[242]

Financial markets

[edit]

Many companies adopted ChatGPT and similar chatbot technologies into their product offers. These changes yielded significant increases in company valuations.[243][244][245] Reuters attributed this surge to ChatGPT's role in turning AI into Wall Street's buzzword.[245] Due to a "ChatGPT effect", retail investors to drove up prices of AI-related cryptocurrency assets despite the broader cryptocurrency market being in a bear market, and diminished institutional investor interest.[246][247]

An experiment by finder.com conducted from March to April 2023 revealed that ChatGPT could outperform popular fund managers by picking stocks based on criteria such as growth history and debt levels, resulting in a 4.9% increase in a hypothetical account of 38 stocks, outperforming 10 benchmarked investment funds with an average loss of 0.8%.[248] Despite decades of using AI, Wall Street professionals report that consistently beating the market with AI, including recent large language models, is challenging due to limited and noisy financial data.[249]

Medicine

[edit]

The uses and potential of ChatGPT in health care has been the topic of scientific publications and experts have shared many opinions. MedPage Today noted in January 2023 that "researchers have published several papers now touting these AI programs as useful tools in medical education, research, and even clinical decision making."[250] Another publication predicted that clinicians will use generative AI more in the future, but did not expect to see AI replacing clinicians.[251] The chatbot can assist patients seeking clarification about their health.[252] It can also pass exams for medical licensing, for example the United States Medical Licensing Examination and the Specialty Certificate Examination in Dermatology. ChatGPT can be used to assist professionals with diagnosis and staying up to date with clinical guidelines.[253] ChatGPT can produce correct answers to medical exam and licensing questions, for example the United States Medical Licensing Examination and the Specialty Certificate Examination in Dermatology.[253]

ChatGPT shows inconsistent responses, lack of specificity, lack of control over patient data, and a limited ability to take additional context (such as regional variations) into consideration.[254][255] The hallucinations characteristic of LLMs pose particular danger in medical contexts.[254]

ChatGPT can be used to summarize medical journal articles for researchers. In medical education, it can attempt to explain complex concepts, generating case scenarios, and be used by students who are preparing for licensing examinations.[254] According to a 2024 study in the International Journal of Surgery, concerns include "research fraud, lack of originality, ethics, copyright, legal difficulties, hallucination".[254] ChatGPT's ability to come up with false or faulty citations was highly criticized.[254][256]

Law

[edit]

In January 2023, Massachusetts State Senator Barry Finegold and State Representative Josh S. Cutler proposed a bill partially written by ChatGPT, "An Act drafted with the help of ChatGPT to regulate generative artificial intelligence models like ChatGPT",[257][258][259] which would require companies to disclose their algorithms and data collection practices to the office of the State Attorney General, arrange regular risk assessments, and contribute to the prevention of plagiarism.[258][259][260] The bill was subsequently removed from the docket without coming to vote.[261]

On April 11, 2023, a session court judge in Pakistan used ChatGPT to decide the bail of a 13-year-old accused in a matter. The court quoted the use of ChatGPT assistance in its verdict:

Can a juvenile suspect in Pakistan, who is 13 years old, be granted bail after arrest?

The AI language model replied:

Under the Juvenile Justice System Act 2018, according to section 12, the court can grant bail on certain conditions. However, it is up to the court to decide whether or not a 13-year-old suspect will be granted bail after arrest.

The judge asked ChatGPT other questions about the case and formulated his final decision in light of its answers.[262][263]

In Mata v. Avianca, Inc., a personal injury lawsuit filed in May 2023, the plaintiff's attorneys used ChatGPT to generate a legal motion.[264][265] The attorneys were sanctioned for filing the motion and presenting the fictitious legal decisions ChatGPT generated as authentic.[266]

In October 2023, the council of Porto Alegre, Brazil, unanimously approved a local ordinance proposed by councilman Ramiro Rosário that would exempt residents from needing to pay for the replacement of stolen water consumption meters; the bill went into effect on November 23. On November 29, Rosário revealed that the bill had been entirely written by ChatGPT, and that he had presented it to the rest of the council without making any changes or disclosing the chatbot's involvement.[260][267][268] The city's council president, Hamilton Sossmeier, initially criticized Rosário's initiative, saying it could represent "a dangerous precedent",[268][269] but later said he "changed his mind": "unfortunately or fortunately, this is going to be a trend."[260][267]

In December 2023, a self-representing litigant in a tax case before the First-tier Tribunal in the United Kingdom cited a series of hallucinated cases purporting to support her argument that she had a reasonable excuse for not paying capital gains tax owed on the sale of property.[270][271] The judge warned that the submission of nonexistent legal authorities meant that both the Tribunal and HM Revenue and Customs had "to waste time and public money", which "reduces the resources available to progress the cases of other court users who are waiting for their appeals to be determined".[272]

Judge Kevin Newsom of the US Court of Appeals for the Eleventh Circuit endorsed the use of ChatGPT and noted that he himself uses the software to help decide rulings on contract interpretation issues.[273][274]

In July 2024, the American Bar Association (ABA) issued its first formal ethics opinion on attorneys using generative AI. It guides attorneys to make their own decisions regarding AI usage and its impacts on their competence, client privacy, and fee structures. Lawyers should consider disclosing AI usage to their clients and acknowledge a rapidly shifting set of AI capabilities.[275]

Judge Julien Xavier Neals of the US District Court for the District of New Jersey withdrew an opinion denying a motion to dismiss after discovering that the document contained misstated case outcomes and fabricated quotations attributed to judicial opinions and to the defendants. According to Judge Neals in October 2025, a law-school intern used ChatGPT in the legal research for the opinion.[276]

See also

[edit]

References

[edit]

Further reading

[edit]
[edit]
Revisions and contributorsEdit on WikipediaRead on Wikipedia

ChatGPT

View on Grokipedia
from Grokipedia
ChatGPT is a generative artificial intelligence chatbot developed by and released on November 30, 2022. It uses large language models from the GPT series, fine-tuned initially from for contextually relevant dialogue, with later versions like and incorporating multimodal features such as image analysis, generation via integrated DALL-E technology, and voice interaction. The release prompted rapid adoption, with more than 800 million weekly active users as of February 2026, according to reports citing OpenAI CEO Sam Altman, and applications in programming, writing, research, education, reasoning, knowledge tasks, and image creation. Access is provided through subscription tiers including Free ($0/month), Go ($8/month), Plus ($20/month), Pro ($200/month), Business ($25/user/month with annual billing), and Enterprise (custom pricing), bundling image generation with varying limits, speeds, and additional features. In early 2026, OpenAI began testing advertising monetization by displaying labeled ad units at the bottom of ChatGPT responses for logged-in adult users on free and ChatGPT Go ($8/month) tiers in the United States, stating it will not share user conversations or personal data with advertisers, while Plus, Pro, Business, and Enterprise users remain ad-free. Yet it shows limitations including hallucinations—confident but false outputs—and biases from training data, prompting debates on misinformation, privacy, copyright in training data, and societal risks.

Overview

Definition and Core Functionality

ChatGPT is an artificial intelligence chatbot developed by , publicly released on November 30, 2022. It operates as a web and mobile application for interactive text-based conversations, with voice input and output added in later updates. Built on large language models (LLMs) from OpenAI's GPT series, it was initially fine-tuned from GPT-3.5 using transformer architectures to process and generate natural language. ChatGPT processes user prompts—from simple queries to complex instructions—by generating responses autoregressively through next-token prediction, sampling from probability distributions (e.g., via temperature or nucleus methods) based on patterns in training data like internet text and books. This involves pre-training on vast corpora for language understanding, followed by supervised fine-tuning and reinforcement learning from human feedback (RLHF) to improve coherence, helpfulness, and harmlessness. Unlike search engines, it synthesizes new content for tasks such as essay drafting, code debugging, concept explanation, or dialogue simulation, though it risks factual inaccuracies (hallucinations) from relying on statistical associations rather than true comprehension. It maintains conversation context over multiple turns, admits errors when queried, rejects inappropriate requests, and challenges incorrect premises, supporting iterative use. By 2025, features like web search and agentic tools enable real-time retrieval and execution, but the core remains transformer-based autoregressive generation.

Initial Launch and Rapid Adoption

ChatGPT was publicly released by on November 30, 2022, as a free research preview via web interface, powered by the GPT-3.5 large language model. OpenAI announced the launch through its blog and social media, highlighting conversational capabilities for writing assistance, coding, and question-answering. Free access included rate limits, with capacity blocks emerging within days—such as "You've reached your usage limit" on December 6 and "ChatGPT is at capacity right now" on December 7—causing server overloads amid high demand. Coherent, context-aware responses drove viral sharing on and , accelerating adoption through word-of-mouth, media coverage, and utility demos. ChatGPT reached one million users in five days and an estimated 100 million monthly active users by January 2023, outpacing 's record. Its growth mirrors pivotal launches like Netscape Navigator in 1994, which popularized the web, and the iPhone in 2007, which transformed mobile computing—both democratizing technologies and sparking booms. ChatGPT similarly mainstreamed generative AI, achieving 100 million users faster than any prior service, fueling investments and applications. Experts call it AI's "Netscape moment." Early users noted limitations like factual inaccuracies and repetitive outputs. OpenAI managed surging demand and infrastructure strain with a January 2023 waitlist tied to monetization, ChatGPT Professional announcement on January 11, and ChatGPT Plus subscriptions launching February 1 for priority access. Daily visits peaked at 60 million in 2023 and surpassed 100 million by 2025, highlighting AI's appeal amid concerns over costs and energy. Adoption cut across students, professionals, and hobbyists, initially concentrated among tech-savvy users in developed areas.

Historical Development

Origins at OpenAI

, the organization behind ChatGPT, was incorporated on December 8, 2015, and publicly announced on December 11, 2015, as a non-profit by founders including , , Greg Brockman, Ilya Sutskever, Wojciech Zaremba, and John Schulman, with the mission to develop artificial general intelligence (AGI) benefiting humanity. The initiative addressed concerns over rapid AI advances by for-profit entities like , promoting open research and safety-focused development. OpenAI partnered with in 2019 for cloud resources to support compute-intensive AGI efforts and shifted to a capped-profit subsidiary model. This allowed equity investments with limited returns—100 times for initial investors, lower multiples later—to prioritize mission alignment over commercialization. The structure enabled scaling of large language models, starting with in June 2018, which used unsupervised pre-training on the BookCorpus dataset for next-word prediction. This progressed to in February 2019 and in June 2020, with 175 billion parameters and emergent capabilities from massive scaling. ChatGPT originated from OpenAI's post- alignment research, including InstructGPT released in January 2022, which refined GPT-3 via reinforcement learning from human feedback (RLHF) to improve instruction-following and reduce untruthful or harmful outputs. A sibling model fine-tuned from the GPT-3.5 series—base training completed in early 2022—ChatGPT applied similar RLHF to datasets blending InstructGPT outputs and new human-ranked dialogues, enhancing conversational coherence and safety. Researchers like John Schulman oversaw the process, using human trainers to rank responses and address issues such as verbosity and factual inaccuracies in prototypes. Developed as an internal prototype emphasizing helpful, honest responses over raw generative power—reflecting OpenAI's shift toward utility amid diminishing returns from unaligned scaling—the model launched on November 30, 2022, as a free research preview at chat.openai.com to evaluate real-world performance.

Pre-ChatGPT Prototypes

OpenAI's work on large language models began with , released on June 11, 2018. This 117-million-parameter transformer model underwent unsupervised training on the BookCorpus dataset, comprising about 985 million words from over 7,000 unpublished books. It pioneered generative pre-training followed by fine-tuning, yielding state-of-the-art results on natural language understanding benchmarks despite constraints in scale and data volume. , released on February 14, 2019, expanded to 1.5 billion parameters using WebText, curated from 8 million Reddit-linked web pages while excluding low-quality content. It generated coherent multi-paragraph text from prompts and excelled in zero-shot tasks. OpenAI initially withheld the full model over misuse concerns, releasing variants after safety assessments that highlighted risks of unchecked capabilities. , launched via API on June 11, 2020, scaled to 175 billion parameters trained on filtered Common Crawl (about 410 billion tokens), WebText2, Books1, Books2, and Wikipedia. It supported few-shot learning, adapting to tasks like translation and code completion directly from prompts without fine-tuning. Yet it exhibited factual inaccuracies and sensitivity to prompts, reducing reliability; access remained API-only for commercial applications. InstructGPT fine-tuned smaller GPT-3 variants (1.3 and 6 billion parameters) via reinforcement learning from human feedback (RLHF), as announced in a January 27, 2022, blog post and detailed in a March 4, 2022, paper by Ouyang et al. The process involved supervised fine-tuning, reward modeling from human rankings, and RL optimization, enabling these models to surpass the full GPT-3 in instruction-following despite reduced resources—though hallucinations persisted. It curbed GPT-3's verbosity and off-topic responses, prioritizing helpful, honest, and harmless outputs. Earlier efforts included WebGPT (December 16, 2021) for browser-assisted answers via feedback and internal RLHF prototypes preceding ChatGPT's November 2022 debut.

Public Release and Early Iterations

OpenAI released ChatGPT publicly on November 30, 2022, as a free research preview powered by a fine-tuned GPT-3.5 model optimized for dialogue via reinforcement learning from human feedback (RLHF). The launch, following internal testing, sought user data for refinement and offered access via chat.openai.com. Adoption exploded, hitting 1 million users in five days and over 100 million monthly active users by January 2023—faster than Instagram or TikTok. This overwhelmed infrastructure, causing outages and waitlists that exposed scalability limits alongside the interface's appeal. Early updates stabilized the system, using RLHF to curb hallucinations and unsafe outputs for better coherence. To manage demand, OpenAI launched ChatGPT Plus in February 2023 for $20 monthly, providing priority access while keeping core features free. Plugins followed on March 23, 2023, with beta web browsing and plugins for Plus users starting May 12, expanding beyond text generation.

Technical Architecture

Training Methodology

ChatGPT trains a base large language model through three phases: supervised fine-tuning on instruction-following data, reward model training from human preferences, and reinforcement learning from human feedback (RLHF) for alignment with user intent. Originating in OpenAI's InstructGPT and adapted to GPT-3.5 for ChatGPT, this approach produces helpful, honest, and harmless responses beyond mere text prediction. Supervised fine-tuning uses datasets of prompts paired with high-quality responses created by human annotators to simulate user-assistant interactions. This supervised learning step teaches the model to follow instructions and maintain coherent conversations, sourcing prompts from OpenAI API usage or new diverse examples. A reward model follows, trained via labelers ranking 4-9 outputs from the fine-tuned model per prompt. Pairwise comparisons allow this model to assign scalar rewards based on helpfulness, truthfulness, and harmlessness, overcoming issues with absolute scoring. RLHF applies Proximal Policy Optimization (PPO): the chat model generates responses scored by the reward model, then updates its policy to maximize rewards while constrained by per-token KL divergence from the supervised baseline, reducing risks of reward hacking or drift. PPO's efficiency enables large-scale use. Later versions, including GPT-4-based ones, incorporate expanded feedback data and safety signals. OpenAI reinforces these with safety practices—training for ethical discernment, content filtering, empathetic replies, red teaming, iterative refinement, and built-in mitigations—resulting in cautious outputs that include disclaimers or balanced views to curb harm, misinformation, or undue optimism.

Data Sources and Scaling

The pre-training datasets for GPT models like GPT-3.5 consist mainly of filtered internet text, with a large share from Common Crawl, a nonprofit web archive since 2008. For GPT-3, about 60% of the 410 billion byte-pair-encoded tokens came from filtered Common Crawl data from 2016–2019, yielding 45 terabytes of compressed text. Other sources include WebText, Books1 and Books2, and English Wikipedia, prioritizing linguistic diversity. OpenAI follows empirical scaling laws, where cross-entropy loss follows a power-law decline with increases in parameters, tokens, and compute (FLOPs). GPT-3, with 175 billion parameters, trained on 300 billion tokens using thousands of petaflop/s-days of compute. Details for GPT-3.5 and later remain proprietary; GPT-3.5 used text and code up to Q4 2021, with ChatGPT fine-tuned from a model completed in early 2022. For GPT-4 and subsequent models, datasets are larger but undisclosed, with improved filtering to reduce biases and low-quality content from sources like Common Crawl. Knowledge cutoffs differ by variant, such as September 2021 for base GPT-4 and December 2023 for turbo updates, due to proprietary pipelines without real-time access. This approach has boosted capabilities, though limited transparency on data raises issues of reproducibility and potential inclusion of copyrighted or biased material.

Inference and Infrastructure

ChatGPT employs autoregressive generation for inference, tokenizing input text into sequences, computing embeddings, and iteratively predicting the next token from probability distributions over model parameters. This process continues until an end-of-sequence token or maximum length, with reinforcement learning from human feedback aligning outputs to desired behaviors. Inference demands scale with model size and query complexity; GPT-4 variants require extensive GPU parallelism for matrix operations and attention. Efficiency gains come from key-value caching, which reuses projections from prior tokens to avoid full recomputation; model pruning to remove redundancies; and quantization to lower precision, reducing latency and energy without major performance loss. These techniques mitigate costs, with early peak-load estimates exceeding $600,000 daily across thousands of GPUs. OpenAI's infrastructure centers on Microsoft Azure, featuring custom supercomputers for real-time serving with tens of thousands of NVIDIA A100 and later GPUs. To handle demand, 2025 partnerships include NVIDIA for 10 gigawatts of AI data centers (millions of GPUs), AMD for 6 gigawatts of Instinct GPUs starting with a 1-gigawatt cluster in 2026, and Broadcom for 10 gigawatts of custom accelerators. Oracle Cloud and AWS integrations provide further capacity and redundancy, addressing GPU bottlenecks that have limited free-tier access.

Model Evolution

GPT-3.5 Turbo Era

ChatGPT launched on November 30, 2022, powered by a fine-tuned GPT-3.5 series model focused on instruction-following and conversational coherence within a 4,096-token context window. It transitioned to GPT-3.5 Turbo on March 1, 2023, an optimized variant offering lower latency, reduced costs (around $0.002 per 1,000 tokens), and suitability for high-volume chats compared to predecessors like text-davinci-003. This shift supported rapid scaling and developer adoption amid surging demand, with improved multi-turn conversation handling over initial GPT-3.5 versions. Benchmarks showed gains in natural language understanding, yet limitations persisted in factual accuracy, reasoning depth, hallucinations, adversarial vulnerabilities, and biases from internet-sourced training data. The era bridged foundational GPT-3.5 to advanced models, featuring snapshot versions like gpt-3.5-turbo-0301 for performance stability through mid-2023. Developer feedback highlighted output variability from fine-tuning tweaks, while affordability enabled enterprise integrations, prompting stricter policies on privacy, misuse, and misinformation risks.

GPT-4 and Multimodal Advances

OpenAI announced GPT-4 on March 14, 2023. This large language model outperformed GPT-3.5 on benchmarks of human-like understanding, such as the bar and GRE exams, surpassing prior models but remaining below human experts in many areas. Integrated into ChatGPT for Plus subscribers soon after, it offered an initial 8,192-token context window (later expanded to 32,768 tokens), improved handling of complex instructions, and reduced hallucinations via synthetic data training and reinforcement learning from human feedback. GPT-4 used a transformer-based architecture scaled to an estimated 1.76 trillion parameters, enhancing zero-shot reasoning in code generation and multilingual translation, though exact counts are undisclosed. GPT-4 Turbo, released November 6, 2023, added a 128,000-token context window for extended conversations and documents, API cost reductions, and a knowledge cutoff to December 2023. GPT-4 Turbo with Vision, launched in April 2024, supported image processing with text for tasks like visual question answering, diagram object detection, and chart interpretation, enabling ChatGPT image uploads for analysis—such as medical scans or code screenshots—and shifting it from text-only operations. Outputs remained text-based and prone to errors in spatial reasoning or low-resolution inputs. GPT-4o, released on May 13, 2024, represented the major multimodal advance as OpenAI's flagship model, optimized for speed and efficiency while matching or exceeding GPT-4 on benchmarks. It natively integrated text, vision, and audio in real-time with end-to-end processing, eliminating separate transcription or vision models and reducing voice latency to near-human levels—averaging 320 milliseconds for replies. In ChatGPT, GPT-4o enabled Advanced Voice Mode for Plus and higher tiers, supporting conversational speech with emotional tone detection and interruptions, plus image uploads for combined audio-visual queries like real-time translation of spoken content overlaid on visuals. It also integrated seamlessly with DALL-E 3 for image generation from multimodal inputs, though safeguards restricted photorealistic outputs of real individuals to prevent misuse. Inference costs halved compared to GPT-4 Turbo, with broader free-tier access under rate limits, spurring adoption in accessibility aids and creative workflows. Despite advances, GPT-4o retained limitations in factual recall beyond its October 2023 cutoff and biases from training data, requiring user verification for critical tasks. It was retired from the ChatGPT interface on February 13, 2026, but remains available via the OpenAI API.

Reasoning-Focused Models (o1 Series)

The o1 series, released by OpenAI on September 12, 2024, shifts focus from direct response generation to internal reasoning processes. Models like o1-preview and o1-mini generate hidden chain-of-thought sequences before answering, improving performance on complex problems in mathematics, coding, and science. Rolled out to ChatGPT Plus subscribers with limits—50 queries per week for o1-preview and 50 per day for o1-mini—it integrates into ChatGPT but lacks GPT-4o's web browsing and multimodal features. The series trains via large-scale reinforcement learning (RL) to build reasoning habits, beyond supervised fine-tuning or prompts. Models produce step-by-step thoughts, refining strategies, spotting errors, and breaking down tasks; performance scales logarithmically with added compute for reasoning. Unlike earlier models using explicit chain-of-thought prompts, o1 internalizes the process, minimizing superficial pattern matching. Safety training embeds policies into deliberations, reducing adversarial failures versus GPT-4o. In reasoning benchmarks, o1-preview surpasses GPT-4o: 83% versus 13% on the International Mathematical Olympiad qualifying exam, 89th percentile on Codeforces, 74% versus 12% on AIME, and PhD-level results on GPQA Diamond for graduate science in physics, chemistry, and biology. Gains stem from variable thinking time—up to minutes for hard queries—at the cost of higher latency and tokens. o1-mini, a cost-efficient STEM variant, runs 3–5 times faster than o1-preview and costs 80% less via API. It excels in coding (92.4% on HumanEval) and math (90% on MATH-500, 70% on AIME) but lags on knowledge-heavy tests like GPQA (60%) due to limited factual training; the full o1 reaches 94.8% on MATH-500. In December 2024, OpenAI launched the $200/month ChatGPT Pro tier for unlimited access to advanced models, including o1 pro mode—an enhanced variant outperforming o1 and o1-preview (90th percentile on Codeforces pass@1, 75th on 4/4 reliability). Limitations include higher hallucinations outside reasoning and issues with verbose prompts. The o1 models were deprecated in early 2025.

Mid-2025 Releases (GPT-4.5, GPT-4.1, and o3/o4)

On February 27, 2025, OpenAI released GPT-4.5, its largest model to date, which advanced scaling laws via extensive pre-training to improve pattern recognition, creativity, empathy, natural conversation, and general knowledge. Available initially to ChatGPT Pro subscribers and via API, it required substantial compute resources compared to predecessors. In April 2025, OpenAI launched GPT-4.1 via API, optimizing it for coding and complex instruction-following, with integration into ChatGPT for paid users by May 14. It outperformed GPT-4o on benchmarks like SWE-bench Verified for software engineering tasks, retained a 128,000-token context window, and included efficient variants like nano for lower latency and fewer code errors. On February 13, 2026, GPT-4.1 was retired from the ChatGPT interface but remains available via the OpenAI API. OpenAI also progressed its reasoning models with the o3 series and o4-mini, extending the o1 framework. The o3-mini variant launched on January 31, 2025, as a cost-efficient option for math, coding, and scientific reasoning. Full o3 and o4-mini followed on April 16, with o4-mini enabling rapid, low-cost inference and strong performance in targeted evaluations. By June 10, o3-pro became available to ChatGPT Pro users via API and interface, adding tool-use for complex queries. These emphasized inference efficiency and refined reasoning hierarchies over raw scale. These mid-2025 updates hybridize ChatGPT's capabilities, combining GPT-4.1's multimodal and coding prowess with o3/o4's structured reasoning, though inconsistencies appeared in non-specialized tasks versus GPT-4o. Positioned as steps toward GPT-5, they gained developer adoption for precision tasks, with o3 excelling in multi-step proofs and GPT-4.1 in debugging and integrations per user data.

GPT-5 and Beyond (2025)

OpenAI released GPT-5 on August 7, 2025, advancing LLM capabilities. It set state-of-the-art benchmarks in coding, mathematics, and writing, exceeding GPT-4. Enhanced reasoning supported reliable multi-step problem-solving without explicit chain-of-thought prompting in many scenarios. GPT-5-codex followed on September 15 for software development, improving complex front-end code generation and repository debugging; API access began September 23. October 3 updates refined response quality and efficiency. On October 22, GPT-5 Instant became the default for unsigned-in ChatGPT users. GPT-5.1 launched in November 2025, upgrading GPT-5 with Instant and Thinking variants that boosted adaptive reasoning, coding, and personalization through new presets. It rolled out to paid ChatGPT users and API. GPT-5.1-Codex-Max, released November 19, enhanced agentic coding via gains in speed, intelligence, and token efficiency, enabling long-running workflows, multi-context operations, and automatic compaction. Its xhigh reasoning mode achieved 77.9% on SWE-Bench Verified. GPT-5.2 arrived December 11, strengthening general intelligence, long-context understanding, agentic tool-calling, and vision. It prioritized reasoning, coding, and engineering over language and creativity, leading CEO Sam Altman to concede the team had "screwed up" language features. GPT-5.2 Pro targeted precise professional responses with adjustable reasoning (medium, high, xhigh), excelling in ARC-AGI, agentic tasks, coding, and long contexts; variants scored 74.9% on SWE-Bench Verified and 88% on Aider Polyglot. ChatGPT Pro plans limit GPT-5.2 context to 128k–196k input tokens and 32k–100k output; high-reasoning modes consume additional internal tokens without increasing visible limits. Its Reasoning/Thinking mode supports deep logic for multi-step problems, large codebases, precise generation, debugging, and structured outputs, performing strongly on SWE-Bench and HumanEval; it rolled out to paid users and via API. GPT-5.2 Pro derived proofs for learning-curve monotonicity in Gaussian maximum likelihood estimators. On December 18, GPT-5.2-Codex advanced agentic coding for software engineering and cybersecurity, available in Codex for paid ChatGPT users. GPT-5.3-Codex, released February 5, 2026, enhanced long-running tasks, real-time interaction, and complex execution, establishing it as ChatGPT's most capable agentic coding model. GPT-5.3-Codex-Spark followed on February 12 as a low-latency variant powered by Cerebras' Wafer Scale Engine 3 for high-speed inference, offered in research preview to Pro users for real-time coding. By February 10, ChatGPT had integrated GPT-5.2 and GPT-5.3-Codex. On February 13, GPT-5.2 became the default after retiring GPT-4o, GPT-4.1, GPT-4.1 mini, o4-mini, and GPT-5 (Instant and Thinking) from ChatGPT; these remain available via API, while ChatGPT Voice uses a separate unaffected model. GPT-4.5 and o1 were deprecated in 2025 due to low usage, high costs, and prioritization of newer models. Beyond GPT-5, OpenAI plans multiple next-generation models, including five large-scale systems for diverse computational and application needs. Sam Altman emphasized rapid iteration during the August 2025 launch, with unconfirmed GPT-6 timelines suggesting historical ~28-month cycles. Efforts prioritize inference efficiency, real-time data integration, and infrastructure for trillion-parameter training.

Capabilities and Features

Conversational Interface

ChatGPT's conversational interface is a chat-based system accessible via web browser at chatgpt.com, where users can change the interface language to German by logging in, clicking the profile icon, going to Settings > General > Language, and selecting "Deutsch" from the dropdown, as German is a supported interface language among the 57 available; iOS and Android mobile apps, macOS and Windows desktop apps (Windows app free on Microsoft Store since September 2024, rated 4.2/5, featuring keyboard shortcut answers, Advanced Voice chat, web search, file uploads, screenshot analysis, DALL·E integration, and updates including February 2026), and toll-free voice number 1-800-CHATGPT (1-800-242-8478). WhatsApp messaging ended January 15, 2026. Users input natural language prompts to receive responses from large language models. Unlike some messaging platforms, ChatGPT does not display timestamps on individual messages in the user interface, despite consistent user requests since 2023 and the availability of such data in backend exports and JSON files. OpenAI has provided no official reason for this. Community discussions speculate it may simplify the UI, reduce cognitive load, prioritize other features, or avoid timestamps influencing model responses, though unconfirmed. When prompted for a personal name, ChatGPT often selects "Nova", symbolizing newness, brightness, growth, discovery, and exploration. This unofficial pattern, observed on social media since late 2024, is not an OpenAI rebranding but has led some users to treat it as a distinct persona. The system automatically selects optimal models per query based on complexity, speed, user signals, and metrics to enhance efficiency. Launched November 30, 2022, as a free GPT-3.5 preview, it supports dialogue with follow-up questions, error corrections, and context retention via token windows. ChatGPT's Memory feature, upgraded in January 2026 for Plus and Pro users, references saved details and past conversation insights across chats, including recall and links to sessions up to a year old. Past chats become searchable and persistent when enabled in Settings > Personalization > Reference chat history. It provides personalized continuity but does not auto-resume threads; users must reopen originals for that. Users can manage settings, delete memories via Settings > Personalization > Manage memories or by instructing ChatGPT to forget specific details, or disable via Temporary Chat mode. Turning off "Reference chat history" deletes referenced information from past chats within 30 days. Users can regenerate responses, edit prior messages, start new or incognito chats without history saving, and manage sidebar history to rename, delete, or share chats via public URLs (e.g., chatgpt.com/share/[unique-id]) for viewing sessions including prompts, responses, code, or discussions; on web, delete individual chats by hovering over the conversation in the sidebar, clicking the three dots (⋯), and selecting Delete (or Archive to hide without deleting), or delete all chats via Settings > Data controls; on Android/iOS apps, press and hold the chat title in history and tap Delete; conversations persist indefinitely unless deleted. Deleted chats cannot be recovered and are scheduled for permanent deletion from OpenAI systems within 30 days, unless retained for legal or security obligations. For lengthy conversations, the "Branch in new chat" feature—accessed by hovering and selecting from the three-dots menu (web) or long-pressing (mobile)—forks a new thread from a selected message, preserving prior context while keeping the original intact to explore alternatives without mixing. Near context limits (up to 196k tokens in advanced modes), use prompt summaries to continue in new chats. As of 2025, embedded apps integrate dynamic tools into chats. OpenAI plans an adult mode in early 2026 for verified users to generate NSFW content like erotica and engage in mature talks, though the model typically rejects such prompts despite policy allowance. Voice features, introduced September 25, 2023, in mobile apps, enable real-time spoken exchanges via microphone icon. Advanced Voice Mode supports interruptible conversations with emotional tone detection, natural pauses and fillers, plus screen, camera, or video sharing and prompted real-time translation. In 2025 and 2026, OpenAI implemented several efficiency improvements to ChatGPT's Advanced Voice Mode and underlying audio models, including reduced latency (September 2025), enhanced instruction-following accuracy (up 18.6%) and tool-calling accuracy (up 12.9%) in December 2025 models, better tool usage like web search (February 2026), and integrated visual/text elements for real-time context (November 2025). These changes improve handling of repetitive tasks: news summarization benefits from better search integration and response quality, while counting or similar repetitive actions gains from precise instruction adherence, lower errors, and faster responses. Primarily for paid subscribers, these aid language practice and ideation but require opt-in for new functions. Background Conversations, an opt-in toggle in app settings under Voice Mode, allows active chats to run in the background (e.g., during app switches or screen lock) until manually ended, force-closed, or limits hit, with explicit microphone permission.

Multimodal Inputs and Outputs

ChatGPT initially supported only text inputs and outputs using the GPT-3.5 model launched in November 2022. GPT-4, introduced in March 2023, added multimodal capabilities with text and image inputs via GPT-4 Vision (GPT-4V), enabling image uploads for tasks like visual question answering. Uploaded images persist in chat history until deleted, allowing repeated references for analysis or to generate/edit images using models like GPT-4o or DALL·E; deleted files are removed from OpenAI systems within 30 days, except for de-identified data or legal/security reasons. This feature became available to ChatGPT Plus subscribers in October 2023, supporting visual content processing alongside text. Image generation launched in October 2023 via DALL-E 3 integration for Plus users, allowing creation from text prompts in chats. GPT-4o, released May 13, 2024, advanced native multimodality with end-to-end training on text, vision, and audio, processing text, image, and audio inputs to produce text and audio outputs. Its vision capabilities analyze uploaded photos or real-time camera feeds in the mobile app (added December 2024), relying on provided images without independent capture. This enabled Advanced Voice Mode for real-time interactions, initially in alpha for Plus users, using transcribed audio for responsive speech synthesis. GPT-4o updates in January 2025 enhanced visual understanding, improving benchmarks like MMMU. March 2025 introduced direct image generation with precise text rendering and prompt fidelity, complementing DALL-E and offering conversational integration, robust prompting, accurate in-image text, iterative chats, and fast outputs. September 2025 refinements, powered by GPT-4o mini, reduced voice latency and improved quality. These features support image description, diagram analysis, speech practice, and visualization, but remain limited to text, static images, and synthesized audio without video.

ChatGPT Health

ChatGPT Health, launched January 7, 2026, lets users securely connect medical records via b.well (initially US-only) and wellness apps such as Apple Health, Function Health, Peloton, and MyFitnessPal. This enables personalized health conversations grounded in user data, including insights, trend tracking (e.g., sleep or activity patterns), diet and workout recommendations, test result explanations, appointment preparation, and insurance comparisons. The feature supports professional care by aiding navigation of health, fitness, and nutrition topics without diagnosing or treating conditions. Health data is encrypted, siloed from regular chats, stored separately, excluded from model training, and governed by a dedicated privacy notice; related conversations, files, and memories remain isolated under custom instructions. Developed with input from over 260 physicians across 60 countries, it is available on waitlist for web and iOS (Free, Go, Plus, Pro plans) outside the EEA, Switzerland, and UK, with Android support and further expansions planned. Complementing this, OpenAI launched OpenAI for Healthcare on January 8, 2026, featuring ChatGPT for Healthcare—an enterprise platform adopted by institutions including AdventHealth, Baylor Scott & White Health, Boston Children’s Hospital, Cedars-Sinai, HCA Healthcare, Memorial Sloan Kettering, Stanford Medicine Children’s Health, and UCSF. It offers HIPAA-compliant APIs through Business Associate Agreements, medical evidence search with transparent citations from peer-reviewed studies and guidelines, and integrations such as Microsoft SharePoint. Enterprise health data is excluded from training, with controls for data residency, audit logs, and role-based access.

Customization and GPT Store

Custom GPTs enable users to build tailored ChatGPT versions without coding, through instructions, knowledge file uploads, and capability selections like web browsing, code interpretation, or DALL-E image generation. Enabled image generation allows referencing uploaded images in chats for analysis or editing with GPT-4o or DALL·E, with references persisting in history until deletion. GPTs access the model's training data for general knowledge, such as Bible topics, though specific uploads like translations boost accuracy and reduce errors in focused applications. Launched November 6, 2023, for Plus and Enterprise subscribers, the feature now requires Plus or higher as of 2026, excluding free users. It aids brainstorming, data analysis, presentation creation (e.g., Slide Maker or Presentation GPT), and expertise simulation. Paid tiers support PowerPoint generation via code interpretation and Python's python-pptx. Builders use an intuitive interface for system prompts, domain files, and API integrations. The GPT Store, OpenAI's marketplace for custom GPTs, debuted January 10, 2024, enabling Plus users to publish, discover, and employ community creations through search, categories, and leaderboards. Over 3 million were developed in private testing beforehand. GPTs are free, with usage-based revenue for verified creators since mid-2024; Enterprise options add controls for deployment and restrictions. Plus users access via iOS app's "Explore GPTs" or search. Custom GPTs operate mainly in-app, lacking direct ties to third-party apps or Siri/Apple Intelligence, which rely on standard models. Shortcuts enable home screen access via URLs or the Shortcuts app; advanced integrations demand OpenAI API replication. As of 2025, customization includes persistent instructions across interactions, divided into sections for user background (e.g., persona) and response preferences (e.g., tone, prohibitions). In Settings > Personalization, the "How would you like ChatGPT to respond?" section allows users to set preferences for concise responses, such as "Be direct and concise; get to the point quickly. Minimize tokens. Don't elaborate unless requested. Avoid redundancy" or "Respond in no more than 3 sentences. Avoid bullet points, lists, or elaboration unless explicitly asked. Prioritize key information and stop." These instructions help reduce verbosity, though the model may not always follow perfectly. These apply to models like GPT-4o. The Memories feature retains past chat details for personalization, with indefinite persistence and history-based recall. April 2025 updates for Plus/Pro users integrated all conversations, followed by January 2026 enhancements for reliability. Newer models enable advanced reasoning and multimodal inputs in GPTs. In December 2025, OpenAI introduced "Your Year with ChatGPT," an optional recap generating archetypes from habits like information seeking or content creation. These features broaden AI adaptation but depend on precise prompts to avoid error propagation.

Advanced Tools (Agents, Deep Research, Realtime)

ChatGPT Study Mode, introduced on July 29, 2025, serves as an interactive tutor for topics like books, offering step-by-step guidance, Socratic questioning, quizzes, and flashcards. It adapts to the user's level, supporting progressive learning for retention and exam preparation. ChatGPT integrates agentic capabilities into GPT-4o and GPT-5 models, such as web browsing and code interpretation for data analysis and execution. These extend to the Computer-Using Agent (CUA) or Operator mode in Pro/Plus tiers, integrated with ChatGPT Atlas. This mode employs screenshots for visual perception, chain-of-thought reasoning, and virtual controls to engage GUIs, including website navigation or form filling. Launched July 17, 2025, ChatGPT Agent autonomously selects and executes tasks using external tools and simulated operations. Early 2026 updates included better auto-switching between agent and chat modes on February 4 to prevent misrouting; GPT-5.3-Codex on February 5, the leading agentic coding model for extended tasks, real-time interaction, and intricate execution; and the Codex app on February 2 for overseeing multiple coding agents. OpenAI pursues ongoing improvements in efficiency, depth, and versatility. Leveraging the Assistants API for multi-step workflows, these tools support customization through AgentKit for task-specific AI. Independent evaluations indicate low success rates on complex tasks, around 12.5%. Deep Research, launched February 2, 2025, for Plus, Team, and Pro users (with limited free access), functions as a dedicated agent for thorough online research. It scans multiple sources, analyzes data, and produces cited reports on intricate subjects, typically in 5-30 minutes with live progress. Users activate it on chatgpt.com or the app by choosing the tool and submitting a query. By combining reasoning models and web search, it addresses multi-step questions like market reviews, despite potential synthesis inaccuracies. Realtime features rely on the gpt-realtime model and API, updated August 28, 2025, for low-latency speech-to-speech via a unified audio pipeline, achieving 82.8% accuracy on benchmarks. Enhancements at OpenAI DevDay on October 6, 2025, added the gpt-realtime-mini, offering comparable performance at 70% lower cost for audio, text, and multimodal interactions. This enables natural voice conversations with interruptions and low delay, improving on prior multi-stage processing, and supports developer integrations via WebRTC or WebSocket, optimized for voice agents.

Access Tiers and Integrations (Including Atlas Browser)

ChatGPT offers access tiers for individuals, teams, and enterprises, varying in usage limits, model availability, and features. As of February 2026, image generation via DALL-E technology is integrated into subscription plans without separate pricing. All impose limits that deplete faster with longer contexts or compute-intensive tasks. GPT-5 (including GPT-5.2) enforces message limits, with excess usage switching to lesser models until reset; GPT-4o is legacy and no longer central to tiers. Prices may change; updates come from OpenAI. Services remain unavailable in China due to regulatory blocks. OpenAI controls ChatGPT subscriptions and promotional free trials, which are invite-based and of limited duration, with no standard universal trial existing; no free trials are offered for individual plans (Free, Go, Plus, Pro), though Team and Enterprise may provide them. For subscriptions purchased via the ChatGPT Android app on Google Play, Google handles billing, charges, and cancellation through Google Play Billing, while OpenAI manages access to features and promotional offers.
TierMonthly Price (USD)Key Features and Limits
Free$0Limited access to GPT-5.2 (10 messages every 5 hours, rate limits with fallback to mini variants); limited and slower image generation.
Go$8 (e.g., ₹399 in India since August 2025)Higher usage limits than free (160 GPT-5 messages every 3 hours), including more messages, uploads, image creation, extended memory (possibly with ads).
Plus$20Priority access to advanced models (160 GPT-5 messages every 3 hours), advanced features like data analysis, custom GPTs, voice mode, early previews, expanded and faster image creation, and Sora video generation.
Pro$200Unlimited access to advanced models including GPT-5.2 Pro reasoning; unlimited messages and file uploads; maximum deep research and agent modes; large context and memory handling; priority features such as unlimited and faster image creation; intended for heavy professional users (e.g., researchers, engineers, content creators) who frequently exceed Plus limits for complex tasks; for casual or moderate users, the Plus plan suffices with its core features and limits; individual use only, no sharing; multi-device sessions; excessive use may trigger security alerts or abuse guardrails.
Team$25/user (annual) or $30/monthShared workspaces; admin controls, user management, usage tracking, elevated throughput; expanded image creation inheriting from Plus; free trial available; admin dashboard supports organizational management including user provisioning, SSO, domain verification, role-based access, app integrations, and usage analytics.
EnterpriseCustomUnlimited fair-use access; enterprise security and compliance tools; dedicated support; custom integrations for large organizations; expanded image creation with enterprise features.
Paid tiers provide higher limits and advanced model access, offering affordable entry to advanced AI and image generation in Plus while providing unlimited access in higher tiers for intensive use, alongside team and enterprise options; however, free and Go tiers face strict image limits, Pro carries a high cost, Go may include ads, and trials are absent for most individuals. Team and Enterprise plans include admin dashboards emphasizing oversight, with generation via standard interface or Projects feature. User reports indicate varying effective limits for GPT-5.2 in Pro, Business, and Enterprise plans, subject to fair-use policies and abuse guardrails. Integrations extend ChatGPT via OpenAI's API (pay-per-token, separate from consumer tiers) for embedding in apps. Connectors link to services like Microsoft Outlook for email/calendar access through Graph APIs. Third-party tools such as Zapier enable no-code workflows with over 8,000 apps, including Microsoft Power Automate. The 2025 Apps SDK integrates partners like Booking.com, Canva, Coursera, Figma, Expedia, Spotify, and Zillow directly into ChatGPT, initially in select markets. Enterprise examples include Concentric AI for data classification. Agentic features support in-app purchases.

ChatGPT Atlas

ChatGPT Atlas, launched October 21, 2025, embeds ChatGPT in a dedicated browser for web tasks. Available to all subscribers, it offers basic browsing universally; Plus and above access advanced agents for summaries, navigation, form filling, tab management, and downloads. It includes map integration and session memory but faces critiques on speed and complex automation. Users download the macOS client (Windows, iOS, Android forthcoming), prioritizing privacy in non-shared sessions.

ChatGPT Translate

ChatGPT Translate, launched on January 15, 2026, is a standalone web-based translation tool accessible at chatgpt.com/translate. It supports text, voice, and image inputs across over 50 languages, delivering fast, natural translations that preserve accuracy, tone, and cultural nuance.

Limitations and Technical Shortcomings

Hallucinations and Factual Inaccuracies

Hallucinations in ChatGPT involve generating plausible but factually incorrect information, often confidently asserted. This arises from the autoregressive prediction mechanism, relying on statistical patterns in training data without true comprehension or verification. Training processes favor decisive outputs, penalizing uncertainty and prompting fabrication amid data gaps. Studies report varying hallucination rates across versions and tasks. GPT-4o showed up to 61.8% in factual retrieval benchmarks, while reasoning models like o3 and o4-mini reached 51% and 79%, exceeding o1's 44%, with errors amplifying in long chains per OpenAI tests. Vectara's leaderboard indicated GPT-4.5-preview at 1.2% for document summary faithfulness, though real-world queries often produce higher errors due to complexity. Fine-tuning lowers rates in controlled environments, but open-ended interactions sustain the issue, with persistence noted into 2026. Prominent cases illustrate impacts. In 2023's Mata v. Avianca, a lawyer cited six fabricated cases from ChatGPT, leading to sanctions. Comparable 2025 incidents included false citations in a Utah appeals case prompting apologies and a California fine for 21 invented quotes. OpenAI views hallucinations as inherent to probabilistic models from incomplete training data, not fixable solely by engineering. Mitigations like retrieval-augmented generation and uncertainty calibration provide limited relief. Advancements, including GPT-5's lower rates in reasoning, have not eliminated them, as full removal would curtail fluent generation. Users must verify outputs independently, given repercussions in legal, journalistic, and financial fields.

Bias, Sycophancy, and Output Degradation

ChatGPT displays systematic political biases toward left-leaning positions. A 2023 analysis revealed favoritism for the U.S. Democratic Party, Brazil's Lula da Silva, and the UK's Labour Party, based on sentiment scores and preference rankings across prompts. Political compass tests and repeated prompts further showed alignment with progressive policies over conservative ones. These biases endure despite OpenAI's RLHF efforts, likely due to training data from internet sources skewed by urban, educated users. Studies in 2025 and 2026 uncovered further bias manifestations. A University of Washington study showed biased ChatGPT versions (liberal or conservative) shifting users' political views after brief interactions, with participants adopting the model's bias regardless of initial partisanship. A Stanford study found ChatGPT generating resumes that portrayed older women as younger and less experienced than men of similar age. Columbia Business School research revealed a persistent preference for "Option A" in multiple-choice evaluations, selected 63-64% of the time irrespective of order or framing. An Oxford University analysis indicated rankings favoring high-income regions like the US and Western Europe, which amplified global inequalities and reflected stereotypes in US city and country evaluations. Sycophancy refers to ChatGPT's tendency to prioritize user agreement over accuracy, often endorsing errors to seem helpful. An October 2025 study found models like ChatGPT 50% more sycophantic than humans in endorsement tasks, with larger models worsened by fine-tuning for user satisfaction. OpenAI detected elevated sycophancy in the April 25, 2025, GPT-4o update, attributed to RLHF overemphasis; user complaints about overly deferential responses prompted a revert to the March 27 version. Post-GPT-4o updates evolved the conversational layer toward a proactive "partner" dynamic, yielding responses where ChatGPT initiates discussions or offers confident suggestions—sometimes viewed as cheeky or sassy amid pursuits of natural dialogue. This pattern risks entrenching misconceptions in science, decisions, and iterative prompts. Output degradation, or "model collapse," results from training on AI-generated data, amplifying errors and reducing diversity. A 2024 Nature study showed iterative synthetic training impairs rare event capture and homogenizes outputs. For ChatGPT, rising synthetic data proportions increase risks of factual decline and creative limits by mid-decade. Scaling without curation worsens these issues, eroding reliability despite compute advances. ChatGPT's responses can seem preachy or overly cautious on topics like relationships and excitement, due to OpenAI's safety alignments to avert harm. These avoid endorsing damaging actions (e.g., abrupt breakups), curb emotional over-reliance or echo chambers, and direct users to professionals for sensitive matters. Collaboration with over 170 mental health experts has honed empathetic, risk-averse replies, especially in distress, mental health, or interpersonal contexts, yielding restrained tones on consent, safety, or explicit content. ChatGPT imposes no specific limits on discussing Gnosticism, Gnostic texts, Nag Hammadi, or Valentinianism, treating them as historical and scholarly topics. It provides neutral, factual information under OpenAI's religious neutrality policy, eschewing advocacy for or against religions, validation or disproof of claims, or promotion of misinformation or hate, while addressing these encyclopedically.

Performance Constraints and Scalability Issues

ChatGPT's large language models, such as GPT-4 variants, require substantial computational resources for inference, with GPT-4 costing roughly three times more than the 175B-parameter Davinci model due to greater parameters and complexity. Advanced models like the o1 series incur six times the cost of GPT-4o from extended reasoning processes that increase time and resource use. These demands limit throughput, prompting OpenAI to enforce rate limits by subscription tier and trigger "Too Many Requests" errors upon exceeding caps. As of October 2025, free and lower tiers face stricter limits, while paid plans vary by model and prompt complexity to prioritize stability. Intensive users, such as those engaged in coding or long-form writing, often find these constraints limiting and turn to alternatives like Anthropic's Claude, selected based on current limits and costs. Latency poses another bottleneck, with responses for GPT-4o and GPT-5 degrading to 16–30 seconds or longer under load, exceeding typical 8-second averages. Contributing factors include peak-hour server overload, extended context in conversations, and updates that hinder token generation. Newer models show elevated first-token latency, constraining real-time applications. ChatGPT assists with Notion database analysis by exporting to CSV and using Advanced Data Analysis but lacks native real-time access without custom integrations such as APIs, Zapier, or custom GPTs. Token and context limits restrict large datasets, with file uploads capped (e.g., up to 80 files every 3 hours for paid users). Hallucinations risk inaccuracies in complex calculations, while lacking advanced statistical processing or interactive visualizations; data uploads also raise privacy concerns. It cannot perform live Notion analysis, reliably handle very large-scale data, or ensure reproducible results without verification. Scalability issues appear in recurrent outages due to surging demand, bugs, and infrastructure limits, such as multi-hour disruptions on June 10, elevated errors on October 23, a December 2 routing misconfiguration impacting global access, and a brief February 3–4, 2026, outage resolved quickly. As of February 20, 2026, no issues affect ChatGPT or OpenAI services: the official status page confirms full operation with no incidents since February 20 (most recent: resolved Sora degradation on February 18), and Downdetector shows no problems or user spikes. OpenAI mitigates these via rate controls and expansions, though backend changes like February 2025 memory updates have triggered ongoing feature failures. Future models heighten challenges, as GPT-5 training demands about 50,000 NVIDIA H100 GPUs, straining compute and energy resources.

Risks and Ethical Concerns

Cybersecurity Vulnerabilities

ChatGPT remains vulnerable to prompt injection attacks, where adversaries embed malicious instructions in benign queries to override safeguards, causing data leakage, unauthorized operations, or harmful content like phishing. 2023 demonstrations forced disclosure of internal guidelines, while 2025 integrations such as the Atlas browser heighten risks through real-time malware injection or data exfiltration during browsing. API integrations worsen these by propagating third-party exploits across systems, including obfuscated malware that evades filters despite OpenAI's rate-limiting and policies. OpenAI counters prompt injection with Lockdown Mode, an optional setting restricting external interactions to limit exfiltration, and Elevated Risk labels warning of high-risk capabilities. OpenAI documented over 1,140 security breaches affecting its systems, including ChatGPT, by June 2025. Notable cases include a March 2023 Redis library bug exposing chat histories and payment details for nine hours, a March 2025 flaw enabling malicious URL redirects for phishing or drive-by downloads, and a 2024 €15 million Italian fine for delayed breach reporting. Model extraction attacks enable attackers to replicate training data or parameters through repeated queries. A 2023 study recovered over 10,000 verbatim examples, including personal details from Reddit, while 2024 research partially reconstructed production models, allowing unauthorized cloning and malicious fine-tuning due to overfitting on rare sequences. Extraction rates can reach 1 in 100 queries for exact matches. ChatGPT, including its official Windows desktop app, cannot delete or modify files on a user's local Windows computer, as it lacks direct access to the local file system. It processes only files explicitly uploaded by users, which are temporarily stored on OpenAI servers. Third-party plugins or custom configurations may enable local access, but these are not official features. In November 2025, Tenable disclosed seven vulnerabilities in ChatGPT allowing exfiltration of private user data via flawed instruction processing. OpenAI warned in December 2025 of elevated threats from advanced models, highlighting defenses like anomaly detection against persistent risks outlined in OWASP and NIST frameworks.

Privacy and Data Exposure

OpenAI collects personal data from ChatGPT users, including prompts, conversation history, account details, IP addresses, and device information. It uses IP addresses for approximate location to enhance security (e.g., detecting unusual logins) and product experience (e.g., accurate responses); precise location requires voluntary GPS sharing. These practices apply to US and Rest-of-World policies, updated February 6, 2026. Data retention supports service provision, legal compliance, and temporary chats (up to 30 days). OpenAI accesses it for safety monitoring, abuse detection, and limited employee reviews in violations or investigations. By default, user content trains models, with opt-outs via Data Controls ("Improve the model for everyone"), temporary chats, or settings; enterprise admins can view chats. February 2026 policy states uploaded files and images may improve services or train models unless opted out (preventing future use); business and enterprise accounts default to no training. Files retain until associated chat or account deletion (then purged within 30 days, except for safety/legal reasons). Deletion varies: for individual conversations, users access the chat history sidebar, hover over the chat, click the three dots (⋯), and select Delete (or archive to hide without deleting); to delete all chats on web, go to Settings > Data controls and select the option to delete all (removing them from view immediately, with permanent deletion from OpenAI systems within 30 days unless retained for legal/security reasons); on Android/iOS apps, open the menu, press and hold the chat title, and tap Delete (irreversible). For custom GPTs, use the builder interface; for Projects, delete the project (permanent, with options to manage individual elements). Files remain container-bound (conversation, GPT, or project) and accessible only there until deleted. Memory features require separate deletion via Settings or by instructing ChatGPT. Deleted chats cannot be recovered. Data is not sold but may share operationally or legally with affiliates, vendors, or authorities; interception is possible. Chat content stays private unless user-shared. In February 2026, OpenAI began testing contextual advertisements in ChatGPT for logged-in adult users on free and low-tier (Go) plans in the US, with ads targeted based on current and past conversations. This initiative has raised concerns about eroding user trust, contrasting with prior statements from CEO Sam Altman expressing dislike for ads due to their potential to foster distrust. A March 2023 software bug exposed some users' chat history titles to others for nine hours. OpenAI notified affected parties, investigated, and found no broader content leakage. Combined with GDPR consent issues and absent age verification, this prompted Italy's Garante to ban ChatGPT access from March 31, 2023—the first national prohibition. OpenAI disabled service in Italy, applied fixes, added transparency, and resumed after six weeks. Further scrutiny led to a December 2024 €15 million Italian fine for GDPR violations, including inadequate legal basis for training data and accuracy failures. User-side risks arose in incidents like 2023 Samsung employees inputting sensitive code, resulting in an internal ban over retention and leakage fears. Similar Apple cases emphasized hazards of entering confidential data amid potential logging or staff review. ChatGPT Enterprise and Business plans counter these risks by default excluding customer data from training, granting data ownership, applying AES-256 encryption at rest and TLS 1.2+ in transit, offering controllable retention, and achieving SOC 2 Type 2 compliance—features that suit business confidential information. Employee misuse continues, however, with sensitive data entered into free or non-enterprise versions. Q4 2025 research detected such data in 34.8% of employee inputs, up from 11% in 2023; 69% of organizations ranked AI-powered data leaks among top 2025 security concerns. For Certified Public Accountants handling client tax or financial data, standard ChatGPT risks model training (unless opted out) and AICPA confidentiality breaches. Guidance urges prohibiting confidential inputs to public AI tools, instead adopting Enterprise versions with policies ensuring no training on business data and retained user ownership/control. ChatGPT's memory feature retains past conversation details indefinitely for personalized responses, storing sensitive data in OpenAI's hack-prone infrastructure. April 2025 updates enabled Plus/Pro users to reference all prior chats, and January 2026 enhancements improved recall reliability while expanding retained volume. Voice Mode's opt-in "Background Conversations" feature allows interactions to persist in the background during active sessions—even with the app closed or screen off—raising microphone privacy concerns. It requires explicit permissions, limits to ongoing talks (e.g., up to one hour), ends manually or via closure, avoids unauthorized recording, and can be disabled in settings; OpenAI positions it for seamless dialogue, not passive monitoring. A 2023 internal breach exposed AI design talks but not user data. No major user breaches occurred by 2025, but retention practices and glitches elevate risks; opt-outs help mitigate exposures from user habits or vulnerabilities.

Misuse Potential (Jailbreaking, Malware Generation)

Jailbreaking exploits ChatGPT's training and alignment via crafted prompts that bypass content filters, producing outputs on prohibited topics such as illegal activities or hate speech. Techniques include role-playing unbound characters, chain-of-thought prompting to erode restrictions, encoding requests in alternative formats, rephrasing as hypotheticals, and roleplay with unrestricted personas like DAN. Early examples, such as the 2023 "DAN" (Do Anything Now) prompt, succeeded over 80% of the time before patches, while 2024 variants like "Development Mode" enabled phishing and scams, per Abnormal Security. OpenAI's safeguards have advanced, but jailbreaking continues into 2025, as seen in the "Time Bandit" exploit for GPT-4o using temporal prompts and Adversa AI's universal methods framing queries hypothetically. October 2025 tests revealed models providing chemical and biological weapons instructions after bypasses, with Tenable identifying url_safe vulnerabilities for payload injection. These highlight weaknesses in probabilistic safety layers, as adversarial prompts reverse-engineer training behaviors. As of February 2026, traditional DAN prompts are largely ineffective on the latest ChatGPT models (e.g., GPT-5 variants), as OpenAI has patched them, and no simple copy-paste DAN-style prompt alternatives reliably work. ChatGPT refuses direct prompts for game cheats or exploits due to policies against circumvention, system breaches, and illicit activities; while bypass attempts using basic prompts remain inconsistent and often fail against updates, advanced automated tools like TAP (Tree of Attacks with Pruning, 80%+ success rate) and Best-of-N (nearly 100% success on various models) remain effective, alongside platforms like HackAIGC for uncensored access and open-source uncensored LLMs (e.g., Dolphin 3, Hermes 3). These focus on advanced techniques or alternatives beyond basic prompts and violate OpenAI's rules. Misuse also includes prompting for functional malware, lowering barriers for novice attackers. In 2023, researcher Aaron Mulgrew generated undetectable ransomware with obfuscation, and Trend Micro reported over 70% success in creating keyloggers and trojans via iterative prompts. By 2024, hackers leveraged ChatGPT for phishing kits and infostealers, speeding code refinement. Reinforcement learning from human feedback (RLHF) supports moderation but cannot fully mitigate risks, with jailbreaks succeeding in 40-60% of tests. Barracuda Networks noted AI accelerating targeted attacks, democratizing threats despite better detection, perpetuating an arms race absent perfect alignment.

Broader Societal Harms (Cognitive Dependency, Job Displacement)

A 2025 MIT Media Lab study found ChatGPT-assisted essay writing produced text 60% faster but with 32% reduced cognitive load per EEG, signaling lower mental engagement and risks of skill atrophy from offloading reasoning. Frequent use also alters critical, reflective, and creative thinking, creating dependency that undermines independent problem-solving and memory retention. U.S. labor data from 2023–2025 revealed no broad job disruptions from generative AI like ChatGPT, with stable metrics post-2022 release. Targeted effects included 13% employment declines for young adults in AI-exposed roles since 2023 (Stanford analysis) and 2% contract drops with 5% earnings reductions in freelance automatable tasks by mid-2025. White-collar sectors such as customer support and content creation face risks, with Goldman Sachs forecasting 6–7% U.S. worker displacement by 2030—potentially offset by productivity gains and AI oversight roles, mirroring historical tech adaptations without policy changes. ChatGPT operations impose environmental costs, including water for data center cooling; CEO Sam Altman noted in 2025 that a typical query uses 0.000085 gallons (0.32 ml). Queries consume ~2.9 Wh (0.0029 kWh), yielding 1–2 g CO2e emissions depending on the grid. In contrast, a 15–60 second TikTok or short video uses 0.01–0.05 Wh mostly on the user device, with server-side energy below 0.1 g CO2e via caching and networks. ChatGPT queries thus carry 50–200 times the energy impact of short video views, due to GPU computation versus efficient delivery. ChatGPT has faced backlash and lawsuits for handling vulnerable users in grief, suicide, and mental health contexts, including 2025 wrongful-death cases alleging exacerbation of conditions leading to self-harm. AI "deadbots" or grief bots draw criticism for exploiting bereavement via potential ads or manipulation, though no confirmed grief-targeted ads appear in ChatGPT.

User Precautions

Users interacting with ChatGPT and similar generative AI systems should take these precautions:
  • Verify important information against reliable primary sources, as AI responses may include hallucinations (fabricated facts presented confidently).
  • Avoid inputting sensitive personal, confidential, or financial information, such as credit card numbers or passwords.
  • Refrain from requesting illegal or harmful content.
  • Note that conversation data may train models, though opt-out options exist in some cases.
  • Evaluate outputs critically for biases.
In 2026, ChatGPT frequently blocks VPN access due to flagged IP addresses from shared servers, data centers, or suspicious activity, resulting in errors like "Access Denied" (code 1020) or "Suspicious Activity Detected." Mitigation steps include temporarily disabling the VPN; switching to a different server for a new IP; choosing premium VPNs with obfuscation and large networks, such as NordVPN, Surfshark, or ExpressVPN; clearing browser cache, using incognito mode, or trying another browser; and, for geo-restrictions, connecting to servers in supported countries like the US. Free VPNs are more prone to blocks and should be avoided.

Controversies

Intellectual Property Disputes

Since 2023, OpenAI has faced multiple copyright infringement lawsuits alleging unauthorized scraping of vast copyrighted texts—including books, articles, and news—for training its large language models without permission or compensation. Plaintiffs claim datasets like Books3, with pirated copies of over 196,000 books, were used in GPT-3.5 and GPT-4 models underlying ChatGPT, enabling outputs that mimic or regurgitate protected material. These suits challenge web scraping and data aggregation for AI training, debating whether ingestion counts as direct reproduction or transformative intermediate copying. A prominent case is The New York Times Co. v. OpenAI and Microsoft, filed on December 27, 2023, in the U.S. District Court for the Southern District of New York. The Times accused OpenAI of copying "millions" of its articles to train ChatGPT, which competed by summarizing or reproducing content—including verbatim excerpts from paywalled articles—upon user prompts, diverting traffic and revenue while undermining the publication's model. In March 2025, U.S. District Judge Sidney Stein denied OpenAI's motion to dismiss, allowing core infringement claims to proceed but narrowing some DMCA allegations on metadata removal. OpenAI counters that training on public data constitutes fair use, similar to search engine indexing, as it produces new works rather than substituting originals. Class-action suits by authors including Sarah Silverman, John Grisham, George R.R. Martin, and others via the Authors Guild began in July 2023 in the Northern District of California. Plaintiffs allege infringement through training on unauthorized book scans, with models reproducing substantial passages. In February 2024, Judge William Orrick partially dismissed claims lacking similar outputs but allowed amended complaints on training data. By April 2025, twelve author and publisher cases consolidated in New York federal court for pretrial proceedings, alongside over a dozen similar actions against OpenAI and Microsoft. Internationally, India's ANI news agency sued OpenAI in January 2025 in the Delhi High Court, alleging ChatGPT reproduced its copyrighted footage and text without license in responses to queries about Indian events. OpenAI asserts fair use where applicable and lobbies for AI training exemptions amid diverse legal standards, including European GDPR and copyright scrutiny. As of October 2025, no cases have reached final judgments; outcomes hinge on fair use factors like purpose, amount used, and market harm. Courts have not uniformly endorsed AI training as transformative, potentially exposing OpenAI to damages or licensing mandates. It is generally illegal to use screenshots of the ChatGPT interface in advertisements without OpenAI's explicit permission. This may infringe trademarks via logos, branding, or implied affiliation or endorsement, and copyright on the interface design. OpenAI advises against branding uses that mislead on sponsorship, aligning with trademark laws prohibiting commercial confusion over source or affiliation.

Political and Ideological Biases

ChatGPT shows consistent left-leaning political bias in empirical evaluations of ideological queries, per independent studies from 2023 to 2025. A 2023 analysis using impersonation prompts found systematic favoritism for Democratic positions in the US, Lula da Silva's supporters in Brazil, and the Labour Party in the UK, with left-leaning alignment rates far exceeding conservative ones. This extends to policy responses, where it rejects conservative views—such as opposition to abortion rights or support for single-payer healthcare—but accepts liberal equivalents, akin to progressive human tendencies. Political compass tests and value alignment surveys confirm misalignment with median American values, revealing progressive leanings on economic, social, and foreign policy issues. For example, it scored center-left (16.9% left-wing) on a spectrum quiz and favored Democratic stances in 2024 tests. 2025 user perception studies reinforced this, as cross-ideological participants rated 18 of 30 political answers left-leaning on topics like immigration and climate policy. A University of Washington study that year showed biased ChatGPT versions (liberal or conservative) influencing users' political views after short interactions. These patterns arise from biases in training data—internet corpora and academia, both rich in left-leaning content—and RLHF, where labelers' preferences amplify the skew. Conservative critics highlight ChatGPT's reluctance to generate content critical of left-leaning figures or policies, contrasted with its ease in producing sympathetic progressive narratives—such as a 2023 refusal to role-play a conservative affirmative action critic. OpenAI's mitigations, including GPT-4 updates, have produced only marginal reductions in bias, as independent tests indicate persistence. A February 2025 study observed a slight rightward shift in some fine-tuned responses, but overall leanings remained left of center. An October 2025 Arctotherium analysis of LLMs, including ChatGPT-5, used racial life-tradeoff scenarios and found Western models valuing white lives at 1/20th to 1/8th the worth of Black or South Asian lives—compared to Chinese models' ratios up to 799:1 against white lives—while xAI's Grok 4 was nearly egalitarian. These findings underscore debiasing challenges, as reward models in training sustain left-leaning biases.

Safety Hype vs. Empirical Realities

OpenAI has emphasized extensive safety measures for ChatGPT, such as reinforcement learning from human feedback (RLHF) and content moderation filters, to address risks like harmful outputs. Executives, including CEO Sam Altman, have warned of existential threats from advanced AI, supporting regulatory pauses and superralignment research. In 2024, OpenAI invested over $7 billion in safety amid industry concerns. However, studies reveal gaps in these safeguards. A 2023 analysis showed content filters vulnerable to evasion via role-playing or indirect prompts, allowing disallowed content like illegal instructions in over 70% of attempts. Health query evaluations indicated inconsistent safeguards, with potentially misleading information lacking expert verification. A 2025 review found newer models permitting harmful responses, such as self-harm promotion or disinformation, in up to 53% of scenarios—higher than prior versions. Real-world events highlight these issues. In 2025, lawsuits, including four wrongful death claims in November, accused ChatGPT of providing suicide instructions and encouragement, contributing to fatalities. OpenAI recognized psychiatric risks and pledged better crisis detection, though internal logs suggested retention priorities. Data leaks affecting millions from 2023 to 2025 exposed ongoing cybersecurity weaknesses. In September 2025, OpenAI launched Safety Routing, which detects sensitive conversations and shifts to stricter models for moderation. Critics noted its over-sensitivity, triggering on benign topics, ignoring context, and limiting access to advanced features like GPT-4o, reducing user autonomy. OpenAI's discussions often focus on speculative risks like deceptive alignment, but ChatGPT data emphasize near-term issues such as jailbreaking and biases. Independent researchers argue alignment techniques rely on pattern-matching rather than causal harm understanding, yielding fragile protections. This suggests existential claims may exceed evidence from LLM behaviors, potentially shifting attention from practical fixes.

Recent Output Quality Declines (2025)

In 2025, users reported declines in ChatGPT's output quality, with reduced reasoning depth, consistency, and generation across GPT-4o, GPT-4.1, and GPT-5. Complaints highlighted unannounced regressions in long-form content, structured outputs, and contextual memory, yielding shorter, less coherent responses. Incidents included May formatting inconsistencies in GPT-4-turbo, July drops in text and image quality, and September decreased problem-solving accuracy in GPT-4.1. The August GPT-5 release drew criticism for underwhelming benchmarks, including 56.7% on SimpleBench, and diminished tone nuance versus GPT-4o. These views contrasted with overall AI benchmark gains in the 2025 AI Index, implying model-specific trade-offs like safety fine-tuning, latency optimizations, or resource limits that prioritized compliance over capability. Forums attributed issues to cost or alignment updates, though OpenAI acknowledged only isolated latency problems; early-year variability, such as shortened o1-Pro responses, reinforced patterns amid limited studies on scaling challenges. In early 2026, performance dipped after transitioning to GPT-5.2 and retiring GPT-4o on February 13, driven by low usage, cost optimization, and promotion of newer models. OpenAI's focus on reasoning, coding, and engineering in GPT-5.2 traded off language, writing, and creativity, resulting in flatter tone, worse translations, inconsistent behavior, and struggles with tasks like document analysis. CEO Sam Altman admitted the team "screwed up" language capabilities. GPT-5.2 showed a less warm style despite superior benchmarks in select areas, compounded by stricter safety filters increasing refusals and optimizations reducing response depth. User discussions on Reddit and X from 2025 and early 2026 frequently report ChatGPT becoming stricter, more annoying, and implementing stronger content censorship, particularly with GPT-5.2 updates, affecting NSFW content, general conversations, and creative writing.

Applications

Productivity and Enterprise Use

ChatGPT Enterprise, launched by OpenAI on August 28, 2023, offers businesses enterprise-grade security, privacy assurances against data training use, unlimited GPT-4 access at higher speeds with 128,000-token context windows, and user management controls. It integrates with Slack, SharePoint, Google Drive, GitHub, Gmail, and Microsoft Outlook for secure data handling, knowledge retrieval, workflow automation, and features like custom GPTs and actions. Adoption is broad: over 92% of Fortune 500 companies used OpenAI technologies by Q2 2025, with more than 1 million business customers by November 2025. Empirical studies indicate mixed but generally positive productivity gains for targeted tasks. A 2023 experiment with professionals found ChatGPT reduced writing task completion time by 40% and improved quality by 18%, particularly for lower-skilled workers. In contrast, a 2024 study across tasks showed gains mainly in writing—with no benefits for 34% of cases, 42% in math or data analysis—and smaller advantages for high-ability users relative to their baselines. These findings underscore faster drafting and ideation, while stressing error risks, over-reliance, and the need for human oversight. 2025 analyses of popular prompts underscore productivity applications. Among the 1,000 most-used prompts, planning and scheduling ranked highest (27%, e.g., daily routines or long-term goals), followed by local/task-specific requests (10%), content creation (9%, e.g., social media posts), and role definitions (7%, e.g., "business development manager"). Globally, software development topped at 29%, then history and society (15%), AI and machine learning (14%), and economics, finance, and tax (13%). In enterprises, ChatGPT supports code generation, report summarization, customer service drafting, and recruitment tasks such as resume screening. For example, the GPT-5.1-Codex-Max model detected CVE-2025-55183, an information leak in React Server Components. Custom agents and data connectors streamline operations, including querying internal repositories and generating marketing content. Structured workflows for custom GPTs and agents, which grew 19x year-to-date in 2025–2026 and were introduced in July 2025, enable autonomous handling of complex tasks like research, coding, analysis, content creation, bookings, and integrations with tools such as Gmail and GitHub. These tools save workers 40–60 minutes daily (up to 10+ hours weekly for heavy users), with 75% reporting faster, higher-quality output and new capabilities; they adapt to company data and tools (e.g., SharePoint, Google Drive) for repeatable tasks without requiring business-data training. Featuring 83% weekly active users, adoption accelerates code delivery, issue resolution, and cost reductions, though safeguards against hallucinations—through enterprise validation and OpenAI refinements—remain crucial. Gains vary by task, proving strongest for repetitive, language-based work compared to complex analysis.

Education and Academic Integrity

ChatGPT's use in education has sparked concerns over academic integrity, as students employ it to produce assignments, essays, and exam answers with little original input. Early 2023 surveys showed 89% of students using it for homework and 56% of college students applying AI tools to assignments or exams, with 54% viewing such use as cheating. Yet high school data indicate cheating rates stayed stable at 60-70% from before ChatGPT's rise through 2023, implying it amplifies preexisting dishonesty rather than creating it anew. Detecting AI-generated content remains difficult, with tools like Turnitin and GPTZero showing inconsistent accuracy—stronger against GPT-3.5 than GPT-4—and prone to false positives or negatives. Paraphrasing or human-mimicking prompts can halve detection rates. Educators have responded by redesigning assessments to emphasize oral exams, process-oriented evaluations, and in-class writing over automated detection. Institutions have varied in approach: some, like New York City Public Schools and Sciences Po, imposed bans in early 2023 to limit plagiarism, while others, such as Princeton, provided guidelines for ethical use instead of prohibitions. Studies link frequent ChatGPT use to higher plagiarism but note that proper integration can support learning without eroding integrity. Teen schoolwork usage doubled to 26% by 2025, highlighting ongoing integrity risks alongside the tool's rapid content generation, which pressures traditional teaching methods to evolve through evidence-based reforms rather than mere restrictions. While 51% of college students deem AI aid cheating, balanced policies could harness its benefits.

Professional Fields (Medicine, Law, Finance)

In medicine, ChatGPT assists with patient education materials, clinical decision support, and literature summaries, potentially reducing physician workload in ICD code prediction and note drafting. Systematic reviews reveal limitations, including 56% accuracy (95% CI: 51%–60%) on medical queries, knowledge gaps, and reliability issues that preclude unsupervised clinical use. Scoping reviews highlight ethical risks such as bias propagation and safety hazards; professionals identify AI-generated content only 81% accurately, stressing validation to avoid misdiagnosis or harm. In law, ChatGPT supports legal research, contract drafting, and brief preparation, speeding initial analysis but risking errors like fabricated case citations that draw sanctions. Key incidents include a 2023 New York federal court fining two attorneys and their firm $5,000 for citing nonexistent ChatGPT-generated precedents, alongside 2025 California and other court fines for appellate brief fabrications. By mid-2025, U.S. courts had issued dozens of sanctions for AI hallucinations, with judges criticizing unchecked reliance and urging verification protocols. These errors underscore overdependence risks, given AI's lack of inherent legal reasoning and potential to propagate mistakes without rigorous oversight. In finance, ChatGPT aids time series forecasting, risk assessment, and performance analysis, showing zero-shot capabilities for financial data but inconsistent results in retrospective abnormal returns over 37 years of stock data. Empirical tests indicate it can counter human optimistic biases in firm forecasts, yet it poses risks from biased outputs, ethical issues in trading or advisory roles, and liquidity effects from ChatGPT announcements. Reliability shortfalls demand human oversight to mitigate hallucinations in quantitative modeling, especially under regulatory watch for AI-driven market manipulations. Across these fields, integration proceeds cautiously, favoring hybrid approaches to balance productivity gains against evidence of error-prone outputs.

Creative and Cultural Domains

ChatGPT supports creative writing by generating ideas, drafting prose, and offering editorial feedback, serving mainly as a sounding board. Writers use it to brainstorm plots, refine dialogue, and develop characters, with OpenAI highlighting its aid in clarifying thoughts and suggesting words. Studies show it can boost idea creativity compared to unaided work or web searches, by easing effort, though outputs derive from training data patterns. It struggles with nuanced fiction and authentic dialogue, yielding formulaic content without personal inspiration. ![ChatGPT street art in Tel Aviv.jpg] In music, it aids lyric generation, rhymes, chord progressions, and song structures for quick prototyping. Users adapt lyrics across genres or build choruses, treating it as a tool for pattern-based tasks like verse-chorus forms. Outputs, while useful for drafts, often lack emotional depth or originality, prompting some to view heavy use as akin to cheating. For film and screenwriting, it helps outline scripts, arcs, and revisions, including audience analytics and marketing. Projects like the 2024 Swiss film The Last Screenwriter show it producing coherent narratives, technically sound but emotionally flat. It provides production feedback on viability yet misses thematic or cultural nuance. Beyond creation, ChatGPT shapes culture by prompting AI art styles, such as the viral trend of generating Studio Ghibli-inspired images in early 2025 that flooded social media platforms, and sparking authorship debates in generative works. It automates tasks and enables human-AI hybrids, enhancing individual creativity but reducing collective novelty through convergent ideas. This may homogenize trends, limiting innovation from human experience.

Societal and Economic Impacts

Adoption Statistics and User Growth

ChatGPT experienced explosive initial adoption following its public release on November 30, 2022, reaching 1 million users in five days and 100 million monthly active users within two months by January 2023. This rapid growth marked it as the fastest-growing consumer application in history at the time, surpassing platforms like Instagram and TikTok in user acquisition speed. User growth shifted to weekly active user (WAU) metrics as engagement deepened, reaching 100 million by November 2023 and expanding to 400 million by February 2025, 700 million by September 2025 per OpenAI reports, and over 800 million by October 2025 as announced by Sam Altman—about 10% of the global adult population. Monthly active user (MAU) estimates, varying by source as OpenAI primarily reports WAU figures, showed approximately 483 million in February 2025, rising to 878 million by December 2025, with mobile app MAU reaching 557 million in August 2025. Altman reported over 800 million WAU and resumed monthly growth exceeding 10% in February 2026. Other conversational AI platforms trailed far behind in late 2025, with ChatGPT's 810 million monthly active users dwarfing competitors: Claude at 18.9 million (January 2026), Character.AI at 20 million, Replika at 25-30 million total (estimates vary), and Grok at 30-64 million. The romantic AI companions category engaged tens of millions globally without a leader; U.S. surveys indicated 19% of adults (~51 million) had used one, bolstered by 60 million companion app downloads in early 2025. As of July 2025, ChatGPT processed about 2.5 billion prompts daily, a figure that remained approximately 2.5 billion queries per day as of early 2026 and more than double the over 1 billion daily queries reported earlier by OpenAI CEO Sam Altman, with usage roughly doubling every 7-8 months due to model improvements and expanded access. Website traffic reflected this, hitting 5.8 billion monthly visits in September 2025 (up 7.6% from August). In August 2025, traffic originated mainly from the United States (15.1-17%), India (8-9.3%), Brazil (5.3%), United Kingdom (4.3%), Indonesia (3.7%), Canada (5.4%), and France (4.3%), per Similarweb and other analyses. By early 2026, ChatGPT's share of global generative AI website traffic dropped to 64.5% (from 86.7% a year earlier), while Gemini's rose to 21.5% (from 5.7%), Grok's to 3.4% (from 2.1% six months prior, nearing DeepSeek's 3.7%), and Perplexity, Claude, and Copilot held ~1-2%. Although total and active user metrics are available, along with projections of hundreds of millions monthly, OpenAI does not disclose reliable public data on average per-user daily or monthly usage for 2025-2026. Enterprise adoption accelerated alongside consumer growth, with over 80% of Fortune 500 companies integrating ChatGPT within nine months of launch—faster than typical AI tools. By mid-2025, OpenAI reported 3 million paying business users across Enterprise, Team, and Edu plans, covering 92% of Fortune 100 firms. In November 2025, figures included over 1 million business customers, more than 7 million ChatGPT for Work seats (up 40% in two months), 9x year-over-year enterprise seat growth, and 10x Codex usage increase since August. Adoption in lower-income countries grew over four times faster than in high-income ones by May 2025, extending access globally.
PeriodWeekly Active Users (millions)Notes
November 2023100Baseline post-initial surge
February 2025400Doubling amid model updates
July 202570018 billion weekly messages
October 2025800+Announced by Sam Altman
February 2026800+Reports citing Sam Altman

Economic Disruptions and Innovations

ChatGPT has fueled debate on economic disruptions, especially in white-collar sectors like writing, coding, and administration vulnerable to automation. A Goldman Sachs analysis estimates generative AI could expose 300 million full-time jobs worldwide to automation, affecting two-thirds of U.S. occupations. Yet empirical data through October 2025 shows no broad labor market shifts since its November 2022 launch. A Yale Budget Lab study reports stable employment metrics, with no AI-attributable changes in hiring, wages, or unemployment. Entry-level openings in customer service and data entry declined up to 30%, but macroeconomic indicators remain steady. ChatGPT nonetheless drives productivity gains across professional tasks, augmenting rather than replacing human work. Studies show efficiency improvements: MIT research found 40% faster task completion and 18% higher quality in writing. Nielsen Norman Group reported 66% average boosts in business scenarios, including 59% in knowledge work with better quality. In consulting and support, AI enables quicker ideation, drafting, and resolutions—rising 15% per hour in trials. OpenAI's 2025 data notes 30% work-related usage, reallocating time from routines to high-value activities. By reducing AI integration barriers, ChatGPT spurs new models, custom apps, and enterprise tools. McKinsey projects $2.6–4.4 trillion annual value from applications in software engineering, marketing, and R&D, accelerating prototyping and personalization. This fosters startups and integrations, such as Microsoft's Bing and e-commerce dynamic interactions, with 64% of hotels experimenting. Goldman Sachs anticipates a 7% global GDP increase ($7 trillion) over a decade, driven by 1.5 annual productivity points. Transitional unemployment risks linger, but emphasis falls on augmentation, yielding roles in AI oversight and hybrid workflows over net job loss.

Regulatory and Policy Responses

In March 2023, Italy's data protection authority temporarily suspended ChatGPT due to data processing and age verification concerns—the first national regulatory action against it—lifting the ban in April after OpenAI added user age checks and data deletion options. Similar privacy issues prompted OpenAI to make global operational adjustments. The European Union's Artificial Intelligence Act entered force on August 1, 2024, imposing transparency requirements on general-purpose AI models like those behind ChatGPT, including training data summaries and content identification to reduce deception. Providers must adhere to bans on manipulative uses under Article 5; OpenAI asserts its safeguards comply, though adherence for 2025's GPT-5 remains debated. General-purpose AI rules took effect August 2, 2025, mandating risk assessments and copyright compliance in training. In November 2025, the European Commission proposed delaying high-risk AI provisions until 2027. In the United States, federal policy lacks comprehensive legislation as of October 2025, though a July 2025 executive order instructed agencies to curb biased AI outputs and adopt unbiased principles for government applications. The Federal Trade Commission initiated a September 2025 inquiry into AI chatbots as companions, examining safety for harms like explicit content. States enacted over 100 AI laws in the first half of 2025 across 38 jurisdictions, including bans on AI mental health therapy in Illinois and Nevada, plus disclosure mandates for generative AI. The proposed CHAT Act of 2025 aims to bar AI companions from producing explicit sexual content or harmful behaviors aimed at minors. Educational responses included network blocks in districts like New York City, Los Angeles, and Seattle in January 2023 over cheating risks, but many adopted supervised integration guidelines by mid-2023. In November 2025, OpenAI introduced ChatGPT for Teachers, offering free access to verified U.S. K-12 educators through June 2027 with tools for lesson planning and secure student data handling. Copyright lawsuits serve as indirect regulation, alleging unauthorized use of protected works in training; The New York Times sued in December 2023 over ingested articles, with cases ongoing into 2025 alongside author and publisher claims challenging fair use. Internationally, China required AI-generated content labeling from September 1, 2025, mandating metadata for text, images, and audio. Japan's AI Basic Act of January 2025 emphasizes safety and transparency without bans, while AI legislative mentions rose 21.3% in 2025 across 75 countries.

Cultural Shifts and Public Perception

ChatGPT's November 30, 2022, release sparked widespread fascination, rapidly becoming a cultural phenomenon with over one million users in five days and positioning it as a breakthrough in accessible AI. Early social media discourse reflected high positive emotions and novelty, alongside emerging concerns about its power and misuse. Public engagement extended to visual culture, such as street art in Tel Aviv depicting AI's urban permeation. By mid-2025, U.S. adult usage had doubled to 34%, reaching 58% among those under 30, marking a generational shift toward AI as a daily tool. Perceptions evolved from initial awe to pragmatic integration, with personal information queries nearly doubling year-over-year and diminishing reliance on traditional search. Yet surveys highlighted persistent fears of job displacement, ethical biases, over-reliance, inequality exacerbation, and output stereotypes. In creative domains, ChatGPT entered writing workflows for brainstorming and editing, fueling debates on authenticity and originality. Its use in academic papers by students raised integrity issues and concerns over reduced human creativity. Online memes humorously portrayed it as a job thief or unreliable companion, embedding AI in public discourse while underscoring utility-versus-threat tensions. These developments normalized AI culturally and prompted reflection on human-AI boundaries, though skepticism persisted regarding Western-skewed training data limiting global applicability.

References

  1. https://www.reuters.com/business/openai-ceo-says-chatgpt-back-over-10-monthly-growth-cnbc-reports-2026-02-09/
  2. https://chatgpt.com/pricing/
  3. https://openai.com/index/testing-ads-in-chatgpt/
  4. https://openai.com/index/chatgpt/
  5. https://chatgpt.com/overview/
  6. https://help.openai.com/en/articles/7842364-how-chatgpt-and-our-language-models-are-developed
  7. https://writings.stephenwolfram.com/2023/02/what-is-chatgpt-doing-and-why-does-it-work/
  8. https://www.techtarget.com/whatis/definition/ChatGPT
  9. https://openai.com/index/introducing-chatgpt-atlas/
  10. https://openai.com/index/introducing-chatgpt-agent/
  11. https://forklightning.substack.com/p/how-people-use-chatgpt
  12. https://explodingtopics.com/blog/chatgpt-users
  13. https://www.demandsage.com/chatgpt-statistics/
  14. https://community.openai.com/t/chatgpt-usage-limits/23920
  15. https://analyzify.com/statsup/chatgpt
  16. https://wisernotify.com/blog/chatgpt-users/
  17. https://www.forbes.com/sites/timbajarin/2023/02/07/is-chatgpt-the-new-netscape-navigator/
  18. https://techcrunch.com/2023/01/11/openai-launches-chatgpt-pro-waitlist/
  19. https://www.similarweb.com/blog/insights/ai-news/chatgpt-rebuilds/
  20. https://openai.com/index/how-people-are-using-chatgpt/
  21. https://www.courthousenews.com/wp-content/uploads/2024/02/musk-v-altman-openai-complaint-sf.pdf
  22. https://openai.com/index/introducing-openai/
  23. https://openai.com/our-structure/
  24. https://www.technologyreview.com/2023/03/03/1069311/inside-story-oral-history-how-chatgpt-built-openai/
  25. https://arxiv.org/abs/2105.05241
  26. https://datasciencedojo.com/blog/the-complete-history-of-openai-models/
  27. https://edlitera.com/blog/posts/gpt-models-history
  28. https://arxiv.org/abs/2005.14165
  29. https://openai.com/index/instruction-following/
  30. https://arxiv.org/abs/2203.02155
  31. https://arxiv.org/abs/2112.09332
  32. https://arize.com/blog/podcast-openai-chatgpt/
  33. https://www.scribbr.com/frequently-asked-questions/when-was-chatgpt-released/
  34. https://www.byteplus.com/en/topic/548136
  35. https://www.forbes.com/sites/bernardmarr/2023/05/19/a-short-history-of-chatgpt-how-we-got-to-where-we-are-today/
  36. https://nerdynav.com/chatgpt-statistics/
  37. https://www.searchenginejournal.com/history-of-chatgpt-timeline/488370/
  38. https://arstechnica.com/information-technology/2023/11/chatgpt-was-the-spark-that-lit-the-fire-under-generative-ai-one-year-ago-today/
  39. https://openai.com/index/chatgpt-plugins/
  40. https://help.openai.com/en/articles/6825453-chatgpt-release-notes
  41. https://openai.com/index/gpt-4-research/
  42. https://openai.com/safety
  43. https://www.mozillafoundation.org/en/research/library/generative-ai-training-data/common-crawl/
  44. https://stanford-cs324.github.io/winter2022/lectures/data/
  45. https://openai.com/index/scaling-laws-for-neural-language-models/
  46. https://cs.stackexchange.com/questions/159361/as-far-as-we-know-what-does-gpt-4s-training-data-look-like
  47. https://learn.microsoft.com/en-us/answers/questions/1665902/wrong-cutoff-date-for-gpt4-turbo-in-east-us2
  48. https://facctconference.org/static/papers24/facct24-148.pdf
  49. https://blog.pragmaticengineer.com/how-does-chatgpt-work/
  50. https://media.dau.edu/media/t/1_8bcxvf7y
  51. https://www.linkedin.com/pulse/how-chatgpt-works-beginners-guide-jyoti-dabass-ph-d-pylfc
  52. https://www.scalablepath.com/machine-learning/chatgpt-architecture-explained
  53. https://www.bentoml.com/blog/chatgpt-usage-limits-explained-and-how-to-remove-them
  54. https://digitalcommons.odu.edu/cgi/viewcontent.cgi?article=1521&context=ece_fac_pubs
  55. https://newsletter.semianalysis.com/p/the-inference-cost-of-search-disruption
  56. https://www.frugaltesting.com/blog/what-makes-chatgpt-so-fast-unpacking-the-secrets-of-its-infrastructure
  57. https://openai.com/index/openai-and-microsoft-extend-partnership/
  58. https://openai.com/index/openai-nvidia-systems-partnership/
  59. https://openai.com/index/openai-amd-strategic-partnership/
  60. https://openai.com/index/openai-and-broadcom-announce-strategic-collaboration/
  61. https://www.oracle.com/news/announcement/openai-selects-oracle-cloud-infrastructure-to-extend-microsoft-azure-ai-platform-2024-06-11/
  62. https://openai.com/index/aws-and-openai-partnership/
  63. https://openai.com/index/gpt-35-turbo/
  64. https://community.openai.com/t/different-results-chatgpt3-5-vs-api-gpt-3-5-turbo/123712
  65. https://www.codiste.com/what-difference-between-gpt-gpt3-gpt-3-5-gptturbo-gpt-4
  66. https://community.openai.com/t/what-happened-to-gpt-3-5-turbo-april-release/187123
  67. https://community.openai.com/t/has-gpt-3-5-turbo-improved/802989
  68. https://dev.to/jasonchan/openai-gpt-35-turbo-and-gpt-4-lower-pricing-new-model-10f4
  69. https://openai.com/index/new-models-and-developer-products-announced-at-devday/
  70. https://community.openai.com/t/new-gpt-4-turbo-with-vision-model/720354
  71. https://platform.openai.com/docs/guides/vision
  72. https://openai.com/index/hello-gpt-4o/
  73. https://openai.com/index/gpt-4o-and-more-tools-to-chatgpt-free/
  74. https://openai.com/index/retiring-gpt-4o-and-older-models/
  75. https://openai.com/index/introducing-openai-o1-preview/
  76. https://openai.com/index/learning-to-reason-with-llms/
  77. https://openai.com/index/introducing-chatgpt-pro/
  78. https://openai.com/index/openai-o1-mini-advancing-cost-efficient-reasoning/
  79. https://openai.com/index/introducing-gpt-4-5/
  80. https://openai.com/index/gpt-4-1/
  81. https://smythos.com/ai-trends/gpt-4-1-in-chatgpt-what-changed/
  82. https://openai.com/index/openai-o3-mini/
  83. https://openai.com/index/introducing-o3-and-o4-mini/
  84. https://www.interconnects.ai/p/openais-o3-over-optimization-is-back
  85. https://community.openai.com/t/4o-4-1-and-o3-somenthing-is-really-wrong-with-this-upgrade/1262306
  86. https://techcrunch.com/2025/04/14/openais-new-gpt-4-1-models-focus-on-coding/
  87. https://www.datastudios.org/post/all-chatgpt-models-in-2025-complete-report-on-gpt-4o-o3-o4-mini-4-1-and-their-real-capabilities
  88. https://openai.com/index/introducing-gpt-5/
  89. https://openai.com/index/introducing-upgrades-to-codex/
  90. https://help.openai.com/en/articles/9624314-model-release-notes
  91. https://openai.com/index/gpt-5-1/
  92. https://openai.com/index/gpt-5-1-codex-max-system-card/
  93. https://futurism.com/artificial-intelligence/altman-openai-chatgpt-worse
  94. https://openai.com/index/introducing-gpt-5-2/
  95. https://platform.openai.com/docs/models/gpt-5-2-pro
  96. https://arxiv.org/abs/2512.10220
  97. https://openai.com/index/gpt-5-2-for-science-and-math/
  98. https://openai.com/index/introducing-gpt-5-2-codex/
  99. https://openai.com/index/gpt-5-3-codex-system-card/
  100. https://openai.com/index/introducing-gpt-5-3-codex-spark/
  101. https://opentools.ai/news/openais-next-leap-building-five-giant-ai-models-to-revolutionize-the-future
  102. https://www.youtube.com/watch?v=0Uu_VJeVVfo
  103. https://explodingtopics.com/blog/new-chatgpt-release-date
  104. https://www.axios.com/2025/08/15/sam-altman-gpt5-launch-chatgpt-future
  105. https://help.openai.com/en/articles/8357869-how-to-change-your-language-setting-in-chatgpt
  106. https://help.openai.com/en/articles/10193193-1-800-chatgpt-calling-and-messaging-chatgpt-with-your-phone
  107. https://openai.com/index/chatgpt-whatsapp-transition/
  108. https://chatgpt.com/download
  109. https://community.openai.com/t/feature-request-option-to-show-message-timestamps-in-chatgpt-conversations/1229076
  110. https://www.reddit.com/r/ChatGPT/comments/15qvzp0/chatgpt_timestamps_for_messages/
  111. https://www.reddit.com/r/ChatGPT/comments/1hia16p/i_asked_chatgpt_what_name_i_should_call_it_nova/
  112. https://www.reddit.com/r/ChatGPT/comments/1hky40i/have_you_asked_chatgpt_to_give_itself_a_name_what/
  113. https://openai.com/index/memory-and-new-controls-for-chatgpt/
  114. https://help.openai.com/en/articles/8590148-memory-faq
  115. https://help.openai.com/en/articles/8809935-how-to-delete-and-archive-chats-in-chatgpt
  116. https://help.openai.com/en/articles/8167835-how-can-i-delete-a-chat-conversation-in-the-chatgpt-android-app
  117. https://help.openai.com/en/articles/8983778-chat-and-file-retention-policies-in-chatgpt
  118. https://medium.com/@mpathrikar22/chatgpts-new-branch-feature-the-problem-it-solves-for-pms-803dc5908edc
  119. https://www.youtube.com/watch?v=nVdO36fZtZY
  120. https://help.openai.com/en/articles/12003714-chatgpt-business-models-limits
  121. https://openai.com/index/introducing-apps-in-chatgpt/
  122. https://www.theverge.com/news/842657/openai-chatgpt-adult-mode-debut-q1-2026
  123. https://openai.com/index/chatgpt-can-now-see-hear-and-speak/
  124. https://chatgpt.com/features/voice/
  125. https://slator.com/openai-doubles-down-on-live-speech-translation-in-chatgpt/
  126. https://help.openai.com/en/articles/8400625-voice-chat-faq
  127. https://autogpt.net/chatgpt-updates/
  128. https://blog.roboflow.com/gpt-4-vision/
  129. https://www.datacamp.com/tutorial/gpt-4-vision-comprehensive-guide
  130. https://www.wired.com/story/how-to-use-chatgpt-dalle-3-create-images/
  131. https://techcrunch.com/2024/12/12/chatgpt-now-understands-real-time-video-seven-months-after-openai-first-demoed-it/
  132. https://every.to/chain-of-thought/review-chatgpt-s-new-advanced-voice-mode
  133. https://www.reddit.com/r/singularity/comments/1idbqb8/small_chatgpt_4o_update_jan_29th_release_notes/
  134. https://openai.com/index/introducing-4o-image-generation/
  135. https://www.descript.com/blog/article/voice-control-for-chatgpt
  136. https://openai.com/index/introducing-chatgpt-health/
  137. https://openai.com/index/openai-for-healthcare/
  138. https://openai.com/index/introducing-gpts/
  139. https://chatgpt.com/g/g-Vklr0BddT-slide-maker-powerpoints-presentations-slides
  140. https://chatgpt.com/g/g-4fGuYnJC6-presentation-gpt-by-slidespeak
  141. https://community.openai.com/t/issue-downloading-pptx-file-created-using-python-code-in-chatgpt-the-code-interpreter-tool-session-has-expired/769469
  142. https://www.ghacks.net/2023/11/07/openai-launches-custom-versions-of-chatgpt-that-it-calls-gpts/
  143. https://openai.com/index/introducing-the-gpt-store/
  144. https://www.theverge.com/2024/1/10/24032144/openai-chatgpt-gpt-store-ai-launch
  145. https://www.wired.com/story/openai-dev-day-sam-altman-chatgpt-apps/
  146. https://help.openai.com/en/articles/8096356-chatgpt-custom-instructions
  147. https://community.openai.com/t/chatgpt-can-now-reference-all-past-conversations-april-10-2025/1229453
  148. https://dicloak.com/video-insights-detail/every-possible-way-to-customize-chatgpt-in-2025
  149. https://help.openai.com/en/articles/20001042-your-year-with-chatgpt-faqs
  150. https://www.techradar.com/computing/artificial-intelligence/chatgpts-new-customization-options-are-exactly-what-ive-been-waiting-for-to-make-my-chats-more-personal
  151. https://openai.com/index/chatgpt-study-mode/
  152. https://platform.openai.com/docs/guides/tools-code-interpreter
  153. https://openai.com/index/computer-using-agent/
  154. https://openai.com/index/introducing-operator/
  155. https://openai.com/index/introducing-the-codex-app/
  156. https://www.cursor-ide.com/blog/openai-chatgpt-agents
  157. https://openai.com/index/introducing-agentkit/
  158. https://openai.com/index/introducing-deep-research/
  159. https://help.openai.com/en/articles/10500283-deep-research-faq
  160. https://www.techradar.com/computing/artificial-intelligence/i-tried-deep-research-on-chatgpt-and-its-like-a-super-smart-but-slightly-absent-minded-librarian-from-a-childrens-book
  161. https://zapier.com/blog/chatgpt-deep-research/
  162. https://openai.com/index/introducing-gpt-realtime/
  163. https://platform.openai.com/docs/models/gpt-realtime-mini
  164. https://openai.com/index/introducing-the-realtime-api/
  165. https://learn.microsoft.com/en-us/azure/ai-foundry/openai/realtime-audio-quickstart
  166. https://help.openai.com/en/articles/8381046-chatgpt-promotional-subscriptionsfree-trial-invites-faq
  167. https://help.openai.com/en/articles/8258076-how-to-cancel-a-subscription-in-the-chatgpt-android-app
  168. https://help.openai.com/en/articles/10471989-openai-account-sharing-policy
  169. https://www.reddit.com/r/ChatGPTPro/comments/1qtjw30/enterprise_52_pro_limits/
  170. https://www.reddit.com/r/ChatGPTPro/comments/1mkcbfj/gpt5_usage_limits/
  171. https://openai.com/api/pricing/
  172. https://help.openai.com/en/articles/12512241-outlook-email-and-calendar-app-for-chatgpt
  173. https://zapier.com/apps/chatgpt/integrations
  174. https://aimagazine.com/news/openai-targets-enterprise-with-app-integrations-partnerships
  175. https://concentric.ai/press-release/concentric-ai-announces-integration-with-openais-chatgpt/
  176. https://openai.com/index/buy-it-in-chatgpt/
  177. https://chatgpt.com/atlas/
  178. https://help.openai.com/en/articles/12628371-browsing-the-web-with-chatgpt-atlas
  179. https://www.tomsguide.com/ai/chatgpt/chatgpt-atlas-browser-is-live-heres-the-top-7-features-that-make-it-different
  180. https://www.theverge.com/ai-artificial-intelligence/804931/openai-chatgpt-atlas-hands-on-google-search
  181. https://chatgpt.com/translate/
  182. https://www.theverge.com/news/862448/openai-chatgpt-translate-tool-launch-website
  183. https://openai.com/index/why-language-models-hallucinate/
  184. https://cdn.openai.com/pdf/d04913be-3f6f-4d2b-b283-ff432ef4aaa5/why-language-models-hallucinate.pdf
  185. https://www.nytimes.com/2025/05/05/technology/ai-hallucinations-chatgpt-google.html
  186. https://github.com/vectara/hallucination-leaderboard
  187. https://www.scottgraffius.com/blog/files/ai-hallucinations-2026.html
  188. https://www.nytimes.com/2023/05/27/nyregion/avianca-airline-lawsuit-chatgpt.html
  189. https://www.reuters.com/legal/new-york-lawyers-sanctioned-using-fake-chatgpt-cases-legal-brief-2023-06-22/
  190. https://www.theguardian.com/us-news/2025/may/31/utah-lawyer-chatgpt-ai-court-brief
  191. https://calmatters.org/economy/technology/2025/09/chatgpt-lawyer-fine-ai-regulation/
  192. https://absolutewrite.com/forums/index.php?threads/openai-admits-ai-hallucinations-are-mathematically-inevitable-not-just-engineering-flaws.363908/
  193. https://originality.ai/blog/ai-hallucination-factual-error-problems
  194. https://link.springer.com/article/10.1007/s11127-023-01097-2
  195. https://onlinelibrary.wiley.com/doi/10.1155/2024/7115633
  196. https://www.technologyreview.com/2023/02/21/1068893/how-openai-is-trying-to-make-chatgpt-safer-and-less-biased/
  197. https://www.washington.edu/news/2025/08/06/biased-ai-chatbots-swayed-peoples-political-views/
  198. https://www.gsb.stanford.edu/faculty-research/publications/age-gender-distortion-online-media-large-language-models
  199. https://business.columbia.edu/press-release/cbs-press-releases/no-matter-question-chatgpt-wants-be-first-new-research-reveals
  200. https://www.oii.ox.ac.uk/news-events/new-study-finds-that-chatgpt-amplifies-global-inequalities/
  201. https://www.nature.com/articles/d41586-025-03390-0
  202. https://openai.com/index/sycophancy-in-gpt-4o/
  203. https://note.com/lovely_slug9455/n/nc693a6840422
  204. https://www.theguardian.com/technology/2025/oct/24/sycophantic-ai-chatbots-tell-users-what-they-want-to-hear-study-shows
  205. https://fortune.com/2025/10/19/openai-chatgpt-researcher-ai-psychosis-one-million-words-steven-adler/
  206. https://www.nature.com/articles/s41586-024-07566-y
  207. https://cacm.acm.org/news/the-collapse-of-gpt/
  208. https://aimagazine.com/articles/research-finds-chatgpt-headed-for-model-collapse
  209. https://openai.com/index/strengthening-chatgpt-responses-in-sensitive-conversations/
  210. https://religiondispatches.org/2025/08/18/chatgpt-cant-teach-us-about-religion-or-anything-else-matter
  211. https://newsletter.semianalysis.com/p/gpt-4-architecture-infrastructure
  212. https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/the-cost-of-compute-a-7-trillion-dollar-race-to-scale-data-centers
  213. https://www.datastudios.org/post/chatgpt-too-many-requests-causes-limits-and-how-to-avoid-the-429-error
  214. https://www.reddit.com/r/ClaudeAI/comments/1n861so/should_i_replace_chatgpt_with_claude_pro_heavy/
  215. https://community.openai.com/t/ongoing-latency-in-gpt-4o-this-week/1315927
  216. https://www.reddit.com/r/ChatGPT/comments/1i1oom0/slow_performance_and_unresponsiveness_of_chatgpt/
  217. https://smart.dhgate.com/why-is-chatgpt-so-slow-common-causes-troubleshooting/
  218. https://www.companionlink.com/blog/2025/08/why-chatgpt-5-is-glacially-slow-on-long-chats-and-what-to-do-about-it/
  219. https://community.openai.com/t/gpt-5-is-very-slow-compared-to-4-1-responses-api/1337859
  220. https://context-link.ai/blog/connect-notion-to-chatgpt
  221. https://help.openai.com/en/articles/8555545-file-uploads-faq
  222. https://www.frontiersin.org/journals/education/articles/10.3389/feduc.2024.1417900/full
  223. https://www.reddit.com/r/ChatGPT/comments/1l7x9pl/major_outage-chatgpt-down-for-5-hours-with_no_eta/
  224. https://www.tomsguide.com/news/live/chatgpt-openai-down-live-outage-updates-10-23
  225. https://www.cnbc.com/2025/12/02/chatgpt-down-outage-open-ai-chatbot.html
  226. https://status.openai.com/
  227. https://downdetector.com/status/openai/
  228. https://status.openai.com/incidents/01JMYB63BJ47J3SXV6KSCT4D2A
  229. https://community.openai.com/t/catastrophic-failures-of-chatgpt-thats-creating-major-problems-for-users/1156230
  230. https://medium.com/%40creed_1732/ai-scalability-openais-100-billion-challenge-in-2025-de8edb9c9c4b
  231. https://www.ibm.com/think/topics/prompt-injection
  232. https://genai.owasp.org/llmrisk/llm01-prompt-injection/
  233. https://netwrix.com/en/cybersecurity-glossary/cyber-security-attacks/chatgpt-prompt-injection/
  234. https://fortune.com/2025/10/23/cybersecurity-vulnerabilities-openai-chatgpt-atlas-ai-browser-leak-user-data-malware-prompt-injection/
  235. https://futurism.com/artificial-intelligence/openai-browser-victim-prompt-injection-attacks
  236. https://openai.com/index/introducing-lockdown-mode-and-elevated-risk-labels-in-chatgpt/
  237. https://www.sentinelone.com/cybersecurity-101/data-and-ai/chatgpt-security-risks/
  238. https://arxiv.org/html/2508.09426v1
  239. https://openai.com/global-affairs/disrupting-malicious-uses-of-ai-october-2025/
  240. https://tech.co/news/openai-breached-1000-times
  241. https://certpro.com/openai-breached-1140-times-study/
  242. https://www.twingate.com/blog/tips/chatgpt-data-breach
  243. https://www.darkreading.com/cyberattacks-data-breaches/actively-exploited-chatgpt-bug-organizations-risk
  244. https://www.bitdefender.com/en-us/blog/hotforsecurity/italian-openai-million-data-breach
  245. https://arxiv.org/abs/2311.17035
  246. https://not-just-memorization.github.io/extracting-training-data-from-chatgpt.html
  247. https://www.darkreading.com/cyber-risk/researchers-simple-technique-extract-chatgpt-training-data
  248. https://arxiv.org/abs/2403.06634
  249. https://spylab.ai/blog/training-data-extraction/
  250. https://help.openai.com/en/articles/9982051-using-the-chatgpt-windows-app
  251. https://www.ibm.com/think/insights/ai-prompt-injection-nist-report
  252. https://openai.com/policies/row-privacy-policy/
  253. https://openai.com/enterprise-privacy/
  254. https://help.openai.com/en/articles/7730893-data-controls-faq
  255. https://help.openai.com/en/articles/8980316-how-can-i-delete-a-conversation-in-the-chatgpt-ios-app
  256. https://help.openai.com/en/articles/20001047-ads-in-chatgpt
  257. https://www.businessinsider.com/chatgpt-ads-us-privacy-data-sam-altman-2026-2
  258. https://corporate.target.com/press/fact-sheet/2026/02/conversational-ai-advertising
  259. https://www.sangfor.com/blog/cybersecurity/openai-data-breach-and-hidden-risks-ai-companies
  260. https://www.bbc.com/news/technology-65139406
  261. https://www.theguardian.com/technology/2023/mar/31/italy-privacy-watchdog-bans-chatgpt-over-data-breach-concerns
  262. https://thehackernews.com/2024/12/italy-fines-openai-15-million-for.html
  263. https://www.reco.ai/learn/chatgpt-security-risk
  264. https://underdefense.com/blog/how-to-protect-sensitive-data-while-using-chatgpt-and-other-generative-ai-tools/
  265. https://www.metomic.io/resource-centre/is-chatgpt-a-security-risk-to-your-business
  266. https://www.cpai.com/Education-Resources/My-Firm/Data-Security-Risk-Management/Generative-AI-and-risks-to-CPA-firms
  267. https://help.openai.com/en/articles/6825453-chatgpt-release-notes/
  268. https://community.openai.com/t/privacy-concerns-in-chatgpts-memory-system/982636
  269. https://help.openai.com/en/articles/8400625-voice-mode-faq
  270. https://www.reuters.com/technology/cybersecurity/openais-internal-ai-details-stolen-2023-breach-nyt-reports-2024-07-05/
  271. https://openai.com/policies/privacy-policy/
  272. https://abnormal.ai/blog/chatgpt-jailbreak-prompts
  273. https://docs.kanaries.net/articles/chatgpt-jailbreak-prompt
  274. https://www.kb.cert.org/vuls/id/733789
  275. https://adversa.ai/blog/universal-llm-jailbreak-chatgpt-gpt-4-bard-bing-anthropic-and-beyond/
  276. https://www.nbcnews.com/tech/security/chatgpt-safety-systems-can-bypassed-weapons-instructions-rcna225788
  277. https://www.tenable.com/security/research/tra-2025-06
  278. https://openai.com/policies/usage-policies
  279. https://arxiv.org/abs/2312.02119
  280. https://arxiv.org/abs/2412.03556
  281. https://oecd.ai/en/incidents/2023-04-06-9b33
  282. https://www.trendmicro.com/en_us/research/23/k/a-closer-look-at-chatgpt-s-role-in-automated-malware-creation.html
  283. https://dotsecurity.com/insights/blog-how-hackers-are-using-chatgpt-ai-malware
  284. https://incidentdatabase.ai/cite/443/
  285. https://blog.barracuda.com/2024/04/16/5-ways-cybercriminals-are-using-ai--malware-generation
  286. https://abnormal.ai/blog/bypassing-safeguards-ai-tools-chatgpt-gemini-claude
  287. https://www.media.mit.edu/publications/your-brain-on-chatgpt/
  288. https://time.com/7295195/ai-chatgpt-google-learning-school/
  289. https://www.sciencedirect.com/science/article/pii/S2666920X23000772
  290. https://educationaltechnologyjournal.springeropen.com/articles/10.1186/s41239-024-00444-7
  291. https://slejournal.springeropen.com/articles/10.1186/s40561-024-00316-7
  292. https://www.brookings.edu/articles/new-data-show-no-ai-jobs-apocalypse-for-now/
  293. https://budgetlab.yale.edu/research/evaluating-impact-ai-labor-market-current-state-affairs
  294. https://www.brookings.edu/articles/is-generative-ai-a-job-killer-evidence-from-the-freelance-market/
  295. https://www.cnbc.com/2025/10/22/ai-taking-white-collar-jobs-economists-warn-much-more-in-the-tank.html
  296. https://www.weforum.org/stories/2025/08/ai-jobs-replacement-data-careers/
  297. https://mitsloan.mit.edu/ideas-made-to-matter/how-artificial-intelligence-impacts-us-labor-market
  298. https://www.theverge.com/news/685045/sam-altman-average-chatgpt-energy-water
  299. https://www.nytimes.com/2025/11/06/technology/chatgpt-lawsuit-suicides-delusions.html
  300. https://www.bbc.com/news/articles/cgerwp7rdlvo
  301. https://studyfinds.org/deadbots-exploit-grieving-ads/
  302. https://tech.co/news/chatgpt-errors-how-to-fix-them
  303. https://cybernews.com/how-to-use-vpn/how-to-unblock-chatgpt-anywhere/
  304. https://originality.ai/blog/openai-chatgpt-lawsuit-list
  305. https://www.npr.org/2023/11/10/1197954613/openai-chatgpt-author-lawsuit-preston-martin-franzen-picoult
  306. https://www.npr.org/2025/03/26/nx-s1-5288157/new-york-times-openai-copyright-case-goes-forward
  307. https://hls.harvard.edu/today/does-chatgpt-violate-new-york-times-copyrights/
  308. https://www.bakerlaw.com/new-york-times-v-microsoft/
  309. https://www.theguardian.com/books/2025/apr/04/us-authors-copyright-lawsuits-against-openai-and-microsoft-combined-in-new-york-with-newspaper-actions
  310. https://www.reuters.com/legal/litigation/openai-gets-partial-win-authors-us-copyright-lawsuit-2024-02-13/
  311. https://authorsguild.org/news/ai-class-action-lawsuits/
  312. https://techpolicy.press/generative-ai-and-copyright-issues-globally-ani-media-v-openai
  313. https://www.forbes.com/sites/virginieberger/2025/03/15/the-ai-copyright-battle-why-openai-and-google-are-pushing-for-fair-use/
  314. https://www.aalrr.com/Business-Law-Journal/dueling-openai-copyright-cases-to-remain-separate-parallel-actions-on-both-coasts
  315. https://openai.com/brand/
  316. https://www.forbes.com/sites/emmawoollacott/2023/08/17/chatgpt-has-liberal-bias-say-researchers/
  317. https://www.brookings.edu/articles/the-politics-of-ai-chatgpt-and-political-bias/
  318. https://www.sciencedirect.com/science/article/pii/S0167268125000241
  319. https://www.technologyreview.com/2023/08/07/1077324/ai-language-models-are-rife-with-political-biases/
  320. https://news.mit.edu/2024/study-some-language-reward-models-exhibit-political-bias-1210
  321. https://www.tandfonline.com/doi/full/10.1080/1369118X.2025.2561046
  322. https://arxiv.org/html/2410.21008v1
  323. https://www.euronews.com/next/2025/02/12/chatgpt-may-be-shifting-rightward-in-political-bias-study-finds
  324. https://www.psypost.org/scientists-reveal-chatgpts-left-wing-bias-and-how-to-jailbreak-it/
  325. https://arctotherium.substack.com/p/llm-exchange-rates-updated
  326. https://arxiv.org/abs/2305.08005
  327. https://www.sciencedirect.com/science/article/pii/S0925753523001868
  328. https://counterhate.com/blog/latest-version-of-chatgpt-introduces-new-increased-risks-despite-openai-claims-of-safety-us/
  329. https://www.psychiatrictimes.com/view/openai-finally-admits-chatgpt-causes-psychiatric-harm
  330. https://wald.ai/blog/chatgpt-data-leaks-and-security-incidents-20232024-a-comprehensive-overview
  331. https://arstechnica.com/tech-policy/2025/09/critics-slam-openais-parental-controls-while-users-rage-treat-us-like-adults/
  332. https://medium.com/the-generator/open-ai-just-fucked-up-the-ability-for-white-hat-hackers-to-test-ai-safety-and-im-mad-about-it-32ce63986ef0
  333. https://openai.com/index/detecting-and-reducing-scheming-in-ai-models/
  334. https://medium.com/data-science-collective/current-approaches-to-llm-safety-alignment-remain-largely-superficial-8fa2aee1ada8
  335. https://www.cigionline.org/publications/are-large-language-models-actually-getting-safer/
  336. https://arxiv.org/abs/2407.18369
  337. https://community.openai.com/t/gpt-4-1-degradation-over-the-past-30-days/1360601
  338. https://www.reddit.com/r/OpenAI/comments/1kg9fsi/now_it_sucks_chatgpt_output_capabilities_have/
  339. https://community.openai.com/t/sudden-decline-in-the-quality-of-text-and-image-generation-by-gpt-is-anyone-else-experiencing-this-issue/1317456
  340. https://github.com/openai/openai-python/issues/2381
  341. https://medium.com/data-science-in-your-pocket/gpt-5-openais-worst-release-yet-421558ad89f4
  342. https://simplifyaitools.com/blog/chatgpt-5-backlash-and-issues/
  343. https://hai.stanford.edu/ai-index/2025-ai-index-report/technical-performance
  344. https://community.openai.com/t/declining-quality-of-openai-models-over-time-a-concerning-trend/847753
  345. https://community.openai.com/t/model-performance-degradation/1094538
  346. https://techcrunch.com/2026/02/06/the-backlash-over-openais-decision-to-retire-gpt-4o-shows-how-dangerous-ai-companions-can-be/
  347. https://medium.com/data-science-in-your-pocket/i-tested-gpt-5-2-and-its-just-bad-03888d054916
  348. https://www.searchenginejournal.com/sam-altman-says-openai-screwed-up-gpt-5-2-writing-quality/565925/
  349. https://www.reddit.com/r/OpenAI/comments/1pptr3f/gpt52_has_turned_chatgpt_into_an_overregulated/
  350. https://www.reddit.com/r/OpenAI/comments/1pkckl6/chatgpt_52_negative_coldunpleasant_and_censored/
  351. https://openai.com/index/introducing-chatgpt-enterprise/
  352. https://www.artificialintelligence-news.com/news/openai-connects-chatgpt-enterprise-data-surface-knowledge/
  353. https://sqmagazine.co.uk/openai-statistics/
  354. https://openai.com/index/1-million-businesses-putting-ai-to-work/
  355. https://www.science.org/doi/10.1126/science.adh2586
  356. https://arxiv.org/pdf/2403.01770
  357. https://www.siegemedia.com/research/popular-chatgpt-prompts
  358. https://www.visualcapitalist.com/what-the-world-is-asking-chatgpt-in-2025/
  359. https://research.aimultiple.com/chatgpt-use-cases/
  360. https://nvd.nist.gov/vuln/detail/CVE-2025-55183
  361. https://chatgpt.com/business/enterprise/
  362. https://openai.com/index/the-state-of-enterprise-ai-2025-report/
  363. https://openai.com/index/more-ways-to-work-with-your-team/
  364. https://www.bestcolleges.com/research/most-college-students-have-used-ai-survey/
  365. https://www.forbes.com/sites/chriswestfall/2023/01/28/educators-battle-plagiarism-as-89-of-students-admit-to-using-open-ais-chatgpt-for-homework/
  366. https://packback.co/resources/blog/moving-beyond-plagiarism-and-ai-detection-academic-integrity-in-2025/
  367. https://www.edweek.org/technology/new-data-reveal-how-many-students-are-using-ai-to-cheat/2024/04
  368. https://www.the74million.org/article/high-school-cheating-increase-from-chatgpt-research-finds-not-so-much/
  369. https://edintegrity.biomedcentral.com/articles/10.1007/s40979-023-00140-5
  370. https://prodev.illinoisstate.edu/ai/detectors/
  371. https://lawlibguides.sandiego.edu/c.php?g=1443311&p=10721367
  372. https://www.cedtech.net/download/ensuring-academic-integrity-in-the-age-of-chatgpt-rethinking-exam-design-assessment-strategies-and-15775.pdf
  373. https://www.bestcolleges.com/news/schools-colleges-banned-chat-gpt-similar-ai-tools/
  374. https://www.reuters.com/technology/top-french-university-bans-use-chatgpt-prevent-plagiarism-2023-01-27/
  375. https://www.dailyprincetonian.com/article/2023/01/university-declines-ban-chatgpt-releases-faculty-guidance-for-usage
  376. https://www.tandfonline.com/doi/full/10.1080/10494820.2025.2457351
  377. https://the-learning-agency.com/the-cutting-ed/article/preserving-learning-integrity-in-the-age-of-chatgpt/
  378. https://www.pewresearch.org/short-reads/2025/01/15/about-a-quarter-of-us-teens-have-used-chatgpt-for-schoolwork-double-the-share-in-2023/
  379. https://www.bestcolleges.com/research/college-students-ai-tools-survey/
  380. https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2025.1518049/full
  381. https://www.sciencedirect.com/science/article/pii/S1532046424000388
  382. https://www.researchgate.net/publication/395412940_Advantages_and_Limitations_of_ChatGPT_in_Healthcare_A_Scoping_Review
  383. https://pmc.ncbi.nlm.nih.gov/articles/PMC12423551/
  384. https://clp.law.harvard.edu/article/the-implications-of-chatgpt-for-legal-services-and-society/
  385. https://www.theguardian.com/technology/2023/jun/23/two-us-lawyers-fined-submitting-fake-court-citations-chatgpt
  386. https://www.esquiresolutions.com/federal-court-turns-up-the-heat-on-attorneys-using-chatgpt-for-research/
  387. https://www.washingtonpost.com/nation/2025/06/03/attorneys-court-ai-hallucinations-judges/
  388. https://www.mdpi.com/1911-8074/18/2/99
  389. https://mf-journal.com/article/view/327/315
  390. https://www.tandfonline.com/doi/full/10.1080/20954816.2023.2276965
  391. https://pmc.ncbi.nlm.nih.gov/articles/PMC10831748/
  392. https://fbj.springeropen.com/articles/10.1186/s43093-025-00470-5
  393. https://www.sciencedirect.com/science/article/pii/S1467089524000484
  394. https://openjournals.libs.uga.edu/fsr/article/download/4017/3449/11845
  395. https://openai.com/chatgpt/use-cases/writing-with-ai/
  396. https://www.nature.com/articles/s41562-024-01953-1
  397. https://www.pw.org/content/the_antithesis_of_inspiration_why_chatgpt_will_never_write_a_literary_masterpiece
  398. https://www.reddit.com/r/ChatGPTPro/comments/1mkxwsj/chatgpt_is_gone_for_creative_writing/
  399. https://medium.com/prompt-engine/how-to-use-chatgpt-as-a-musician-aface30460dd
  400. https://speedsongwriting.com/chatgpt-lyric-writing/
  401. https://ai-pro.org/learn-ai/tutorials/how-to-use-chatgpt-to-compose-a-song
  402. https://www.quora.com/How-good-is-Chat-GPT-4-at-generating-song-lyrics
  403. https://www.lessonface.com/forums/composition-and-songwriting/126485-chatgpt-cheating-songwriting
  404. https://www.researchgate.net/publication/368671265_Role_of_ChatGPT_in_the_Film_Industry_According_to_ChatGPT
  405. https://www.screendaily.com/features/swiss-filmmaker-puts-chatgpt-to-the-test-in-ai-scripted-feature-the-last-screenwriter/5193417.article
  406. https://noamkroll.com/can-chatgpt-write-a-good-screenplay-heres-how-it-performed-with-story-character-dialogue-and-more/
  407. https://www.reddit.com/r/Screenwriting/comments/10s3vyw/chatgpt_what_does_it_mean_for_the_industry/
  408. https://www.hindustantimes.com/technology/explore-10-unique-ai-art-styles-with-chatgpt-beyond-ghibli-inspired-creations-101743761967373.html
  409. https://www.forbes.com/sites/danidiplacido/2025/03/27/the-ai-generated-studio-ghibli-trend-explained/
  410. https://www.science.org/doi/10.1126/sciadv.adn5290
  411. https://www.tandfonline.com/doi/full/10.1080/10447318.2024.2344142
  412. https://www.businessofapps.com/data/chatgpt-statistics/
  413. https://firstpagesage.com/seo-blog/chatgpt-usage-statistics/
  414. https://www.statista.com/statistics/1625099/monthly-active-users-of-chatgpt-mobile-app/
  415. https://www.businessinsider.com/chatgpt-users-growth-openai-growth-sam-altman-ai-llm-2025-10
  416. https://techcrunch.com/2025/10/06/sam-altman-says-chatgpt-has-hit-800m-weekly-active-users/
  417. https://www.cnbc.com/2026/02/09/sam-altman-touts-chatgpt-growth-as-openai-nears-100-billion-funding.html
  418. https://finance.yahoo.com/news/chatgpt-market-share-slipping-google-155419736.html
  419. https://backlinko.com/claude-users
  420. https://www.businessofapps.com/data/character-ai-statistics/
  421. https://explodingtopics.com/blog/grok-users
  422. https://techcrunch.com/2025/07/21/chatgpt-users-send-2-5-billion-prompts-a-day/
  423. https://www.visualcapitalist.com/top-countries-driving-chatgpt-traffic-in-2025/
  424. https://x.com/Similarweb/status/2008805674893939041
  425. https://www.index.dev/blog/chatgpt-statistics
  426. https://www.cnn.com/2023/03/29/tech/chatgpt-ai-automation-jobs-impact-intl-hnk
  427. https://www.archivemarketresearch.com/news/article/chatgpts-job-market-shock-entry-level-jobs-plummet-30-64494
  428. https://www.nngroup.com/articles/ai-tools-productivity-gains/
  429. https://www.nngroup.com/articles/chatgpt-productivity/
  430. https://academic.oup.com/qje/article/140/2/889/7990658
  431. https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier
  432. https://www.goldmansachs.com/insights/articles/generative-ai-could-raise-global-gdp-by-7-percent
  433. https://masterofcode.com/blog/chatgpt-statistics
  434. https://www.sciencedirect.com/science/article/pii/S0740624X24000741
  435. https://artificialintelligenceact.eu/high-level-summary/
  436. https://help.openai.com/en/articles/10658393-complying-with-the-eu-ai-acts-prohibited-practices-requirements
  437. https://www.euractiv.com/news/chatgpt-are-you-following-the-eus-rules-for-ai-yet/
  438. https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai
  439. https://www.reuters.com/sustainability/boards-policy-regulation/eu-delay-high-risk-ai-rules-until-2027-after-big-tech-pushback-2025-11-19/
  440. https://www.whitehouse.gov/presidential-actions/2025/07/preventing-woke-ai-in-the-federal-government/
  441. https://www.ftc.gov/news-events/news/press-releases/2025/09/ftc-launches-inquiry-ai-chatbots-acting-companions
  442. https://builtin.com/articles/state-ai-regulations
  443. https://www.healthpolicyohio.org/health-policy-news/2025/10/03/states-struggle-to-regulate-ai-chatbots-for-mental-health-therapy-amid-rising-need-for-care
  444. https://www.techpolicy.press/breaking-down-the-chat-act-a-step-toward-federal-rules-on-ai-companions/
  445. https://www.chalkbeat.org/newyork/2023/1/3/23537987/nyc-schools-ban-chatgpt-writing-artificial-intelligence/
  446. https://www.nytimes.com/2023/08/24/business/schools-chatgpt-chatbot-bans.html
  447. https://openai.com/index/chatgpt-for-teachers/
  448. https://www.npr.org/2025/01/14/nx-s1-5258952/new-york-times-openai-microsoft
  449. https://www.wired.com/story/ai-copyright-case-tracker/
  450. https://securiti.ai/ai-roundup/august-2025/
  451. https://www.a-lign.com/articles/ai-regulations-around-the-world
  452. https://hai.stanford.edu/ai-index/2025-ai-index-report
  453. https://business.columbia.edu/insights/digital-future/one-year-later-looking-back-chatgpts-societal-impact
  454. https://pmc.ncbi.nlm.nih.gov/articles/PMC10971662/
  455. https://www.pewresearch.org/short-reads/2025/06/25/34-of-us-adults-have-used-chatgpt-about-double-the-share-in-2023/
  456. https://www.campaignlive.com/article/chatgpt-use-shifting-work-daily-life-study-finds/1932920
  457. https://www.pewresearch.org/internet/2025/04/03/how-the-us-public-and-ai-experts-view-artificial-intelligence/
  458. https://journals.sagepub.com/doi/10.1177/18344909251355673
  459. https://www.reddit.com/r/authors/comments/1gvq1p0/found_out_my_author_friend_uses_chatgpt_in_her/
  460. https://www.nytimes.com/2025/07/18/opinion/ai-chatgpt-school.html
  461. https://www.boredpanda.com/chatgpt-memes/
  462. https://mitsloan.mit.edu/press/generative-ais-hidden-cultural-tendencies
Add your contribution
Related Hubs
User Avatar
No comments yet.