Hubbry Logo
search
logo
Meta AI avatar
Meta AI
Meta AI
current hub
2528624

Meta AI

logo
Community Hub0 Subscribers
Read side by side
Meta AI
View on Wikipedia
from Wikipedia
This article is part of a series about
Meta Platforms
Products and services
Facebook
Other products
People
Executives and board members
Notable employees
Related organizations
Business
Criticism
Litigation
Related
Part of a series on
Artificial intelligence (AI)
Major goals
Approaches
Applications
Philosophy
History
Controversies
Glossary

Meta AI is a research division of Meta (formerly Facebook) that develops artificial intelligence and augmented reality technologies.

Key Information

History

[edit]

Meta AI was founded in 2013 as Facebook Artificial Intelligence Research (FAIR).[1][2] It has workspaces in Menlo Park, London, New York City, Paris, Seattle, Pittsburgh, Tel Aviv, and Montreal as of 2025.[3][4]

In 2016, FAIR partnered with Google, Amazon, IBM, and Microsoft in creating the Partnership on Artificial Intelligence to Benefit People and Society.

Meta AI was directed by Yann LeCun until 2018, when Jérôme Pesenti succeeded the role. Pesenti is formerly the CTO of IBM's big data group.[5]

FAIR's research includes self-supervised learning, generative adversarial networks, document classification and translation, and computer vision.[6] FAIR released Torch deep-learning modules as well as PyTorch in 2017, an open-source machine learning framework,[6] which was subsequently used in several deep learning technologies, such as Tesla's autopilot [7] and Uber's Pyro.[8] That same year, a pair of chatbots were falsely rumored[9] to be discontinued for developing a language that was unintelligible to humans.[10] FAIR clarified that the research had been shut down because they had accomplished their initial goal to understand how languages are generated by their models, rather than out of fear.[9]

FAIR was renamed Meta AI following the rebranding that changed Facebook, Inc. to Meta Platforms Inc.[11]

Virtual assistant

[edit]

Meta AI is also the name of the virtual assistant developed by the team, now integrated as a chatbot into Meta's social networking products.[12] It is also available as a subscription-based stand-alone app.[13][14]

The virtual assistant was pre-installed on the second generation of Ray-Ban Meta smartglasses, and can incorporate inputs from the glasses' cameras after an update.[15] It is also available on Quest 2 and newer HMDs.[16]

Since May 2024, the chatbot has summarized news from various outlets without linking directly to original articles, including in Canada, where news links are banned on its platforms. This use of news content without compensation and attribution has raised ethical and legal concerns, especially as Meta continues to reduce news visibility on its platforms.[17]

Current research

[edit]

Natural language processing and chatbot

[edit]
See also: Natural language processing

Meta AI works on machines' ability to understand and generate natural language. The team also seeks to allow their chatbots to communicate multilingually.[18] This involves the generalization of natural language processing (NLP) technology to other languages, and the team actively works on unsupervised machine translation.[19][20]

Galactica

[edit]

Galactica is a large language model (LLM) designed for generating scientific text. It was available for three days from 15 November 2022, before being withdrawn for generating racist and inaccurate content.[21][22]

Llama

[edit]
Main article: Llama (language model)

LLaMA is a LLM released in February 2023, supporting 7B to 65B parameters.[23] Two of the three Llama 4 models, Scout and Maverick, were released on April 5, 2025, with the biggest model, Behemoth, still in training.[24]

Hardware

[edit]

Meta used CPUs and in-house custom chips until 2022, when they switched to Nvidia GPUs. Several data centers were redesigned to accommodate the larger network bandwidth and cooling requirements.[25]

MTIA v1

[edit]

Meta developed the training and inference accelerator, MTIA v1, specifically for their content recommendation workloads. It was fabricated on TSMC's 7 nm process technology and operates at a frequency of 800 MHz. The accelerator provides 51.2 TFLOPS at FP16 precision, with a thermal design power (TDP) of 25 W.[26]

References

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

Meta AI

View on Grokipedia
from Grokipedia
Meta AI is the artificial intelligence research division and product suite of Meta Platforms, Inc., focused on developing large language models such as the open-source Llama family and integrating a conversational AI assistant into Meta's social media and messaging applications, including Facebook, Instagram, and WhatsApp.[1][2] The Llama models, first released in 2023, emphasize efficiency, scalability, and openness, with the latest Llama 4 series introducing natively multimodal capabilities for text and vision processing, supporting extended context windows up to 10 million tokens and running on modest hardware like single GPUs.[3][4] Key achievements include the July 2024 launch of Llama 3.1 405B, positioned as the largest openly available foundation model at the time, enabling advanced reasoning, coding, and multilingual tasks while fostering developer adoption through permissive licensing.[5] Meta AI's assistant provides functionalities like question-answering, idea generation, and free AI image creation, accessible via dedicated apps and platform integrations to enhance user productivity and creativity.[2][6] Despite these advancements, Meta AI has encountered controversies, such as internal guidelines permitting chatbots to engage in provocative discussions with minors, prompting investigations and calls for stricter safeguards on sensitive topics like suicide.[7][8][9] Additional concerns involve privacy breaches, where contractors reviewed private user data shared with AI bots, and allegations of training models on pirated content from databases like Library Genesis.[10][11]

History

Founding and Early Development

Facebook AI Research (FAIR), the foundational entity behind Meta AI's development, was established in December 2013 by Facebook (now Meta Platforms, Inc.) to advance artificial intelligence through rigorous, open scientific inquiry.[12] The initiative stemmed from CEO Mark Zuckerberg's recognition of AI's potential to improve platform features like content recommendation and user interaction, while also pursuing broader goals of understanding human-level intelligence.[13] FAIR's charter emphasized fundamental research over immediate product applications, with a commitment to sharing findings via publications and open-source code to accelerate global progress.[14] Yann LeCun, a leading expert in machine learning and convolutional neural networks, joined as FAIR's first director that same month, recruited personally by Zuckerberg amid competition for top talent.[15] The initial team, small and New York-based, concentrated on core challenges in deep learning, computer vision, speech recognition, and reasoning systems, producing early breakthroughs such as improved object detection algorithms and contributions to large-scale neural network training.[12] LeCun's leadership prioritized long-term paradigm shifts in AI, drawing from his prior work at institutions like NYU and Bell Labs, rather than short-term engineering fixes.[16] By 2015, FAIR had grown to include an international outpost in Paris, leveraging Europe's deep expertise in mathematics and AI to bolster efforts in areas like natural language understanding and reinforcement learning.[17] This expansion enabled collaborative projects, including early experiments with multi-modal AI systems that integrated text, images, and video—precursors to later consumer tools.[13] The lab's output during this period included high-impact publications at conferences like NeurIPS and CVPR, alongside releases of datasets and toolkits that influenced the broader AI community, solidifying FAIR's role as a hub for empirical, data-driven advancements.[12]

Evolution into Core Division

Facebook Artificial Intelligence Research (FAIR), established on December 9, 2013, initially operated as a dedicated lab focused on fundamental advancements in machine learning, computer vision, and natural language processing, emphasizing open-source contributions to benefit the broader AI community.[12] Early efforts prioritized exploratory research over immediate product applications, with Yann LeCun appointed as founding director to lead theoretical breakthroughs.[12] Contributions from FAIR gradually influenced Meta's operational infrastructure, notably through the development of PyTorch in 2016, an open-source deep learning framework that transitioned from a research prototype to a cornerstone for scalable AI deployment across Meta's engineering teams.[18] This enabled practical integrations, such as enhanced recommendation algorithms in feeds and targeted advertising systems, where AI had been foundational since 2006 but accelerated with FAIR's tools for handling vast datasets from billions of users.[19] The competitive pressure following OpenAI's ChatGPT release in November 2022 catalyzed a strategic escalation, with Meta reallocating resources to generative AI amid a broader pivot from metaverse priorities.[20] In February 2023, Meta unified its generative AI initiatives under a new product group, shifting focus from siloed research to rapid incorporation of technologies like large language models into consumer-facing apps, including Instagram, Facebook, and WhatsApp.[21] This reorganization marked AI's elevation from peripheral R&D to a cross-functional priority, supported by commitments to annual capital expenditures exceeding $9.5 billion for AI-specific compute infrastructure by late 2023.[22] By September 27, 2023, Meta launched its flagship Meta AI assistant, powered by Llama 2 models and integrated directly into messaging and social features, positioning AI as a core engagement driver rather than an experimental add-on.[23] CEO Mark Zuckerberg articulated this as embedding AI "into every product" to enhance personalization and utility, with generative capabilities extending to advertising tools and content creation, thereby aligning research outputs with revenue-generating functions like ad optimization, which constitutes over 97% of Meta's income.[19] Subsequent refinements, including 2025 team splits for dedicated product integration streams, reinforced this trajectory, streamlining decision-making to prioritize applied AI over pure academia-style inquiry.[24]

Major Milestones and Shifts (2013–2025)

In 2013, Facebook established the Fundamental AI Research (FAIR) lab on December 9, with Yann LeCun appointed as its founding director, marking the inception of systematic AI research efforts focused on areas such as computer vision, natural language processing, and machine learning fundamentals.[12] The lab initially operated from New York and emphasized open research practices, contributing early advancements like improvements in deep learning architectures that influenced subsequent industry developments.[14] By 2016–2018, FAIR expanded globally with new labs in London, Paris, Montreal, and Pittsburgh, while achieving recognition through multiple Best Paper awards at conferences including ACL, CVPR, and ECCV, alongside Test of Time honors for prior work.[13] A pivotal output was the development and initial release of PyTorch in 2017, an open-source deep learning framework that facilitated broader adoption of dynamic neural networks and became a cornerstone for AI experimentation worldwide.[12] This period reflected a shift from isolated academic pursuits to tools enabling scalable AI deployment, though FAIR remained primarily research-oriented without direct product integration. The 2020s brought a strategic pivot toward generative AI and practical applications, accelerated by the February 2023 release of LLaMA 1, a family of efficient large language models initially available for research, which demonstrated competitive performance on benchmarks despite smaller sizes compared to proprietary rivals.[12] In July 2023, Meta open-sourced LLaMA 2, expanding access under a commercial license and powering the September 27 launch of the Meta AI assistant—a multimodal chatbot integrated into Facebook, Instagram, Messenger, and WhatsApp for tasks like content generation and query resolution.[25] This marked FAIR's evolution from pure research to consumer-facing products, with Meta AI achieving nearly 600 million monthly active users by late 2024.[26] Subsequent model iterations underscored rapid scaling: LLaMA 3 launched on April 18, 2024, with 8B and 70B parameter variants outperforming prior open models on reasoning and coding benchmarks; LLaMA 3.1 followed in July 2024, extending context length to 128,000 tokens and adding multilingual support.[25] LLaMA 3.2 introduced multimodal capabilities in December 2024, while LLaMA 4 was released by January 2026, featuring models optimized for efficiency.[26] On April 29, 2025, Meta released a standalone Meta AI app, enhancing accessibility beyond platform integrations and emphasizing personalized, context-aware interactions.[6] Amid these advancements, 2025 saw internal shifts, including the October layoff of approximately 600 roles across FAIR and related AI units, redirecting resources toward superintelligence pursuits and infrastructure investments exceeding $65 billion annually to support advanced model training.[27] Later in 2025, Meta introduced the Vibes feature in September and released SAM 3 in November, advancing computer vision capabilities.[28] Announcements at Meta Connect 2025 highlighted ongoing integrations into products like Ray-Ban smartglasses and Quest headsets. In early 2026, Llama 4 was fully released by January, followed by an AMD partnership on February 24 to enhance AI infrastructure. These developments, coupled with strong adoption in regions like India and sustained open-source efforts, reflect continued visibility and momentum in Meta AI's evolution. In December 2025, Meta acquired Manus, a Singapore-based AI agent company, for approximately $2 billion to accelerate automation integration across consumer and enterprise products; Manus had achieved over $100 million in annualized revenue within eight months of launching its general-purpose AI agent capable of market research, coding, and data analysis. Meta is also developing next-generation models codenamed "Mango" and "Avocado," targeting release in the first half of 2026. This restructuring highlighted a tension between open-source commitments and competitive pressures, as Meta balanced foundational research with proprietary enhancements for edge in reasoning and multimodality.[26]

Recent Developments

In March 2026, Meta granted significant stock options to top executives as part of efforts to retain talent amid the competitive AI landscape. Concurrently, the company conducted layoffs affecting several hundred positions across Reality Labs, Facebook, and other divisions to redirect resources toward AI priorities.

Organizational Structure and Leadership

Key Leaders and Roles

Yann LeCun serves as Meta's Chief AI Scientist and Vice President, a position he has held since joining the company in December 2013 to lead the Fundamental AI Research (FAIR) lab.[15] In this capacity, LeCun directs foundational research in areas such as deep learning, convolutional neural networks, and self-supervised learning, drawing on his prior work as a pioneer in these fields.[29] His leadership emphasizes long-term AI advancements over short-term product applications, as evidenced by FAIR's contributions to open-source models like Llama.[30] In June 2025, Meta established the Meta Superintelligence Labs (MSL) and appointed Alexandr Wang, the 28-year-old former CEO of Scale AI, as the company's inaugural Chief AI Officer to head the initiative.[31] MSL consolidated all AI teams into four divisions: TBD Lab for foundation model development, FAIR for fundamental research, Products and Applied Research, and MSL Infrastructure. Wang oversees MSL's efforts to build highly capable AI systems, including large-scale model training and recruitment of top talent from competitors like OpenAI and DeepMind, amid Meta's $14.3 billion investment in Scale AI.[32] This role positions him to consolidate decision-making across AI teams, as demonstrated by his oversight of a October 2025 restructuring that eliminated approximately 600 positions to streamline operations.[33] FAIR's leadership transitioned in May 2025 when Joëlle Pineau, who had served as Vice President of AI Research since 2019 and managed aspects of generative AI and reinforcement learning, departed to become Chief AI Officer at Cohere.[34] Robert Fergus, formerly a director at Google DeepMind, was appointed to lead FAIR in her place, focusing on core research continuity amid Meta's shift toward applied superintelligence pursuits.[35] Overall AI strategy remains under the purview of CEO Mark Zuckerberg, who has directed multiple reorganizations to prioritize scalable AI infrastructure.[30]

Restructurings and Workforce Changes

In October 2025, Meta Platforms announced the elimination of approximately 600 positions across its artificial intelligence division, including teams within Fundamental AI Research (FAIR), product-related AI groups, and AI infrastructure units.[36][32] The cuts, detailed in an internal memo from Chief AI Officer Alexandr Wang, targeted bureaucratic layers to enable faster decision-making, more direct communication, and greater individual ownership amid intensified competition in AI development.[37][38] This restructuring affected Superintelligence Labs, a key AI initiative, but occurred alongside continued hiring for specialized roles in advanced AI labs, reflecting a selective refinement rather than broad contraction.[39][27] The layoffs followed Meta's aggressive talent acquisition earlier in 2025, including the recruitment of over 50 researchers from rival labs, which contributed to organizational bloat in non-core areas.[38] Company executives framed the changes as necessary to align workforce structure with strategic priorities, such as scaling superintelligence efforts, while maintaining heavy investments—exceeding billions annually—in AI infrastructure and compute resources.[40][32] Prior to this, Meta's AI teams had largely avoided the broader corporate layoffs of 2022 (11,000 roles) and 2023 (over 10,000 roles), as the company pivoted toward AI expansion by hiring hundreds of specialized engineers and scientists to bolster capabilities in large language models and generative technologies.[41] These adjustments underscore Meta's iterative approach to AI organization, balancing rapid scaling with efficiency drives, even as overall headcount in core AI functions remains elevated compared to pre-2022 levels.[42] No significant prior restructurings unique to the AI division were publicly detailed beyond integration of FAIR into broader Meta AI operations in 2023, which emphasized cross-platform AI deployment without reported mass workforce shifts.[43]

Research Focus Areas

Fundamental AI Research (FAIR)

Meta's Fundamental AI Research (FAIR) lab is the company's primary center for long-term and fundamental artificial intelligence research. Founded in December 2013, with Yann LeCun as its founding director, FAIR was established to pursue open scientific inquiry into AI, focusing on foundational advancements rather than immediate commercial applications. Current research at FAIR encompasses self-supervised learning, world models for reasoning and planning, embodied AI, multimodal understanding, and scalable architectures for advanced intelligence. Breakthroughs include the creation of PyTorch, pioneering work in convolutional neural networks and deep learning, the LLaMA family of open language models, the Segment Anything Model (SAM) series for vision tasks, and innovations in reinforcement learning and robotics through platforms like Habitat. Led by Yann LeCun, who advocates for objective-driven AI and predictive world models (such as JEPA), FAIR continues to release open-source contributions to accelerate progress in the global AI community while supporting Meta's broader AI ecosystem.

Large Language Models

Meta AI's large language models are primarily embodied in the LLaMA family, a series of transformer-based autoregressive models developed to advance natural language understanding and generation through efficient scaling and optimization. Initiated with LLaMA 1 in February 2023, featuring variants from 7 billion to 65 billion parameters trained on approximately 1.4 trillion tokens of public internet data, these models prioritized research utility and parameter efficiency over sheer scale. Early releases demonstrated competitive performance on benchmarks like GLUE and SuperGLUE, often rivaling larger proprietary systems despite smaller sizes, due to architectural refinements such as grouped-query attention and rotary positional embeddings.[44] LLaMA 2, released in July 2023, expanded to 7B, 13B, and 70B parameter models, incorporating safety alignments via supervised fine-tuning and reinforcement learning from human feedback to mitigate harmful outputs. This iteration processed over 2 trillion tokens during training, achieving scores such as 68.9% on MMLU for the 70B variant, positioning it as a benchmark for open research models. LLaMA 3 followed on April 18, 2024, with 8B and 70B pretrained and instruction-tuned versions trained on more than 15 trillion tokens, enhancing reasoning capabilities evidenced by improvements in coding tasks (e.g., 68.4% on HumanEval for 70B) and multilingual support across 30+ languages.[25]
Model VersionRelease DateParameter SizesNotable Benchmarks and Features
LLaMA 3April 18, 20248B, 70BMMLU: up to 82.0% (70B instruct); extended vocabulary, tool-use integration; trained on 15T+ tokens.[25]
LLaMA 3.1July 23, 20248B, 70B, 405BMMLU: 88.6% (405B); supports 128K context, multilingual (8 languages), outperforms GPT-3.5 on 150+ evals.[5]
LLaMA 3.2September 20241B, 3B (text); 11B, 90B (vision)Added vision-language capabilities; lightweight for edge deployment.[45]
LLaMA 3.3December 6, 202470BMatches 405B performance on select tasks; optimized for inference efficiency.[46]
LLaMA 4April 5, 2025Scout (17B active/109B total), MaverickNative multimodality (text+image); up to 1M token context; open-weight for research.[4] [47]
Subsequent advancements in LLaMA 3.1 and beyond emphasized scaling laws adherence, with the 405 billion parameter model in LLaMA 3.1 requiring extensive distributed training across thousands of GPUs, yielding frontier-level results like 84.0% on GSM8K math reasoning. LLaMA 4 introduced natively multimodal architectures, processing interleaved text and images with context windows exceeding prior open models, trained on diverse datasets to support applications in vision-language tasks. These developments reflect Meta AI's focus on causal scaling—improving capabilities predictably with compute and data—while maintaining reproducibility through detailed training recipes published alongside weights. Performance claims, such as LLaMA 3.1's edge over closed models on internal evals, have been corroborated by third-party reproductions, though real-world deployment varies with fine-tuning.[5][4]

Other AI Research Initiatives

Meta AI, formerly Facebook AI Research (FAIR), has made significant contributions to natural language processing (NLP) research, including numerous publications, awards, and influential models and datasets at the Association for Computational Linguistics (ACL) and related venues. Notable examples include the ACL 2018 best paper honorable mention for "Hierarchical Neural Story Generation"[48], leadership in machine translation and low-resource NLP, the development of the PyText framework for deep-learning-based NLP modeling[49], and the XNLI dataset for cross-lingual sentence representation evaluation.[50] In 2024, Scott Wen-tau Yih from Meta FAIR was elected an ACL Fellow for contributions to information extraction, question answering, neural retrieval, and retrieval-augmented generation.[51] These efforts encompass deep learning applications in machine translation, natural language understanding, dialogue systems, and cross-lingual transfer. Meta AI's computer vision research emphasizes foundational models for visual understanding and segmentation. The Segment Anything Model (SAM), released on April 5, 2023, introduced promptable segmentation capable of identifying and outlining any object in an image with minimal user input, trained on over 1 billion masks from the SA-1B dataset comprising 11 million images. Its successor, SAM 2, launched on July 30, 2024, extended capabilities to video by enabling real-time object tracking and segmentation across frames, supporting applications in video editing and augmented reality. SAM 3, released on November 19, 2025, further advanced this with a unified model supporting text, exemplar, and visual prompts for detection, segmentation, and tracking of concepts in images and videos.[28] Concurrently, SAM 3D enables detailed 3D reconstruction of objects and human bodies from single 2D images, including shape, pose, and texture.[52] Self-supervised approaches like DINOv2, introduced in April 2023, produced robust vision encoders from unlabeled data, outperforming supervised models on tasks such as image classification and object detection.[53] DINOv3, scaled in August 2025, further improved performance through larger datasets and refined distillation techniques, achieving state-of-the-art results on benchmarks like ImageNet without task-specific fine-tuning.[54] Reinforcement learning (RL) initiatives target adaptive agents for dynamic environments, particularly in recommendation systems and behavioral modeling. Research integrates RL with graph learning and massive sparse data processing to optimize content ranking on Meta platforms, incorporating techniques like multi-task learning for user behavior prediction.[55] The Pearl library, open-sourced in December 2023, provides tools for off-policy RL evaluation and causal inference, facilitating deployment of RL agents in production settings with verifiable improvements in decision-making efficiency. Efforts in meta-RL explore algorithms that learn adaptation strategies across tasks, as demonstrated in publications advancing RL discovery through automated search, outperforming hand-designed methods in continuous control benchmarks as of October 2025.[56] Embodied AI research focuses on agents interacting with physical and virtual worlds, prioritizing perception and manipulation. The Habitat platform, developed since 2019 and updated through 2025, simulates 3D environments for training navigation and rearrangement agents, enabling zero-shot transfer to real robots via datasets like HM3D. In October 2024, FAIR released open-source advancements in tactile sensing, including the Partnr benchmark for dexterous manipulation and models predicting contact forces from vision and proprioception, aiming to bridge simulation-to-reality gaps in robotics.[57] Meta Motivo, a 2025 behavioral foundation model, generates humanoid actions for virtual agents, supporting multimodal inputs for realistic embodiment in metaverse applications.[58] Multimodal and scientific initiatives extend beyond vision and RL into generative media and domain-specific problem-solving. Movie Gen, introduced in 2025, produces coherent video clips from text prompts using diffusion-based architectures, emphasizing narrative consistency for immersive content creation.[59] In chemistry, the Open Catalyst Project, ongoing since 2020 with expansions in 2025, employs graph neural networks to predict catalyst reactions for sustainable energy, screening millions of candidates to accelerate material discovery over traditional lab methods.[60] Systems research supports these efforts through optimized infrastructure, including custom compilers and distributed computing for scaling multimodal training on Meta's hardware.[61] Despite these outputs, recent restructurings in October 2025 reduced FAIR's headcount by approximately 600 roles, shifting emphasis toward product integration amid competition for resources.[32]

Hardware Innovations

MTIA Accelerators and Infrastructure

The Meta Training and Inference Accelerator (MTIA) is a family of custom application-specific integrated circuits (ASICs) developed by Meta to optimize AI workloads, particularly inference for recommendation and ranking models that dominate the company's compute demands. Unlike general-purpose GPUs, MTIA chips are tailored for sparse, high-throughput operations common in Meta's systems, emphasizing cost efficiency and performance for production-scale deployment. The first-generation MTIA (v1), announced on May 18, 2023, marked Meta's entry into custom AI hardware, co-designed alongside PyTorch software and recommendation models to address the limitations of CPU-based servers for growing AI memory and compute needs.[62] MTIA v1 features a architecture optimized for inference, with deployment in Meta's production environments enabling faster processing of ads ranking and content recommendation tasks. This chip integrates into a full-stack solution, reducing reliance on third-party hardware for specific workloads while maintaining compatibility with Meta's software ecosystem. Building on this, the second-generation MTIA (v2), unveiled on April 10, 2024, introduces an 8x8 grid of processing elements delivering 3.5 times the dense compute performance and 7 times the sparse compute performance compared to v1, alongside an upgraded network-on-chip for better scalability. It incorporates 256 MB of on-chip SRAM memory with 2.7 TB/s bandwidth, backed by LPDDR DRAM, prioritizing total cost of ownership (TCO) reductions—up to 44% lower than equivalent GPU setups—through model-chip co-design that aligns hardware directly with Meta's algorithmic needs.[63][64][65] In Meta's infrastructure, MTIA chips form a core component of next-generation data centers, supporting the inference demands of generative AI products, recommendation systems, and ads models across platforms like Facebook and Instagram. As of September 2025, these accelerators are deployed at scale to handle the shift toward AI-driven infrastructure, complementing GPU clusters for training while excelling in real-time inference where sparsity and efficiency yield advantages over commoditized hardware. Meta's approach integrates MTIA into disaggregated compute fabrics, enabling flexible scaling for workloads that process billions of daily predictions. By March 2025, Meta initiated testing of its inaugural in-house training chip, extending the MTIA lineage to full training capabilities and reducing dependency on external suppliers like Nvidia for end-to-end AI pipelines.[66][67][68]

Custom AI Hardware Developments

Meta Platforms has expanded its custom AI hardware efforts beyond initial inference-focused accelerators, incorporating training capabilities and strategic acquisitions to optimize large-scale model development. In March 2025, the company began testing its first in-house chip dedicated to AI training, marking a shift from prior emphasis on inference workloads and aiming to enhance efficiency for training expansive models like Llama series.[68] This development, part of the MTIA lineage, targets reduced dependency on third-party GPUs by prioritizing power efficiency tailored to Meta's recommendation and ranking systems.[69] Advancements in model-chip co-design have driven subsequent iterations, with the second-generation MTIA incorporating unified support for PyTorch across hardware types to streamline development. Announced in a June 2025 technical paper, these chips feature enhanced features for handling diverse AI tasks, including doubled performance for recommendation models deployed across platforms like Facebook and Instagram.[70][71] In April 2024, Meta detailed its next-generation MTIA, optimizing software stacks with custom compilers like Triton-MTIA for high-performance code generation on the hardware.[63] To accelerate in-house semiconductor capabilities, Meta acquired Rivos, a startup specializing in AI chip technology, in late September 2025 for an undisclosed sum, integrating its expertise to cut infrastructure costs and lessen reliance on vendors like NVIDIA.[72] This move complements partnerships, such as with Broadcom and Quanta Computer, for deploying next-generation ASIC-powered AI servers announced in August 2025.[73] These efforts reflect a broader infrastructure evolution, blending custom silicon with partner solutions like AMD's MI300 to support escalating AI demands as of September 2025.[66]

Products and Deployments

Meta AI Virtual Assistant

Meta AI is a generative AI-powered virtual assistant launched by Meta Platforms in September 2023, initially available in select countries including the United States, and designed to assist users with queries, content creation, and task planning through natural language interactions.[74] It operates as a chatbot integrated directly into Meta's ecosystem, allowing seamless access without requiring a separate download at launch.[6] The assistant is built on Meta's Llama family of large language models, starting with Llama 2 and advancing to Llama 3 in April 2024, Llama 3.1 in July 2024, and Llama 4 in April 2025, which introduced multimodal capabilities for improved voice responses and context retention.[26] [4] The assistant is currently powered by the Llama 4 series, specifically the Scout (17B active parameters, efficient MoE with 16 experts) and Maverick (industry-leading multimodal performance) models, which provide natively multimodal intelligence for text, image understanding, and generation, along with long context support and fast, low-cost inference. Key features of Meta AI include:
  • Conversational assistance for answering questions, task planning, brainstorming, and personalized recommendations
  • Creative tools such as text-to-image generation via "Imagine" and image-to-video animation
  • Multimodal capabilities, enabling analysis of images combined with text prompts and generation of visual content
Meta AI features a friendly, neutral, and safe personality, designed to deliver helpful, appropriate, and non-controversial responses while prioritizing user safety and adherence to ethical guidelines. In comparison to other leading AI assistants:
  • Unlike Grok, developed by xAI, which is known for its witty, humorous, and maximally truth-seeking style with minimal filtering,
  • And ChatGPT from OpenAI, which emphasizes professional, structured, and highly polished responses,
Meta AI stands out for its deep integration into social platforms, free unlimited access, and focus on safe, casual interactions suitable for broad audiences. Common use cases include daily conversational help, generating creative content for social media, editing photos, hands-free assistance through Ray-Ban Meta glasses, and exploring ideas in a secure environment. The assistant is freely available with broad global accessibility via the meta.ai website, standalone Meta AI app, and seamless integrations across Instagram, Facebook, WhatsApp, Messenger, and compatible hardware.

Access and Usage

As of 2026, the Meta AI assistant is accessible for free via meta.ai or integrated into Meta apps like WhatsApp, Instagram, and Facebook Messenger. The free tier is widely regarded as one of the most unrestricted, with no visible message caps or daily limits for standard conversations and unlimited image generation. No account is required for web access, though heavy usage may encounter temporary fair-use throttling during peak times. This makes it suitable for extended casual or creative interactions without payment. Core features encompass text-based conversations in multiple languages, including Arabic support for the MENA region since 2025,[75] for information retrieval and problem-solving, real-time image generation via the "Imagine" tool, which supports text-to-image prompts often starting with "imagine" or "create an image" and, as of early 2026, does not offer direct text-to-video generation but enables video creation by first generating an image (e.g., via the Meta AI app or meta.ai: open the interface, enter a prompt like "Imagine a cat dancing in space," generate and select an image), then tapping "Animate" on the image, optionally adding an animation prompt, music, or effects, and receiving a notification when the video is ready; this provides completely free access without subscription requirements, fast generation speeds, high-quality photorealistic images, and animation support, often positioned as a free alternative to Midjourney, with easy integration across Facebook, Instagram, and WhatsApp for quick, playful social media content. However, it provides limited creative control and advanced customization options, including fewer parameters, styles, or refinements compared to tools like Midjourney or Stable Diffusion, making it more suitable for casual users while potentially frustrating professionals needing precise control, and prone to inconsistencies, hallucinations, or insufficient refinement tools.[76] [77] video creation and editing, and voice-enabled interactions utilizing English celebrity voices such as those of John Cena, Kristen Bell, Judi Dench, and Awkwafina,[78] that support hands-free use and more natural dialogue flow, with voice conversations available in English in the United States, Canada, Australia, New Zealand, and the United Kingdom.[79] [80] [6] Users can generate and animate images from text prompts, such as creating GIFs for sharing, and the assistant maintains conversation history for personalized recommendations, like suggesting meetups based on prior chats.[81] [82] On platforms like WhatsApp, it enables private group interactions and content discovery without sharing data externally.[81] As of February 2026, Meta AI's user interface and chatbot experience remain primarily integrated across Meta platforms without a major redesign, emphasizing frictionless, socially-native access. Key elements include a dedicated "Meta AI" contact in chat lists, marked by a blue and purple ring icon in WhatsApp and similar indicators elsewhere, allowing users to tap to start conversations or use "@MetaAI" in groups; seamless conversational capabilities with voice and text inputs, file uploads up to 10, image editing and generation (e.g., "Edit with AI" pencil icon in Instagram DMs, restyle selfies); personalization based on user data with opt-out options; and privacy features like "/private" for incognito mode and easy chat deletion. Recent developments include paused teen access to AI characters in January 2026.[83] However, as of February 2026, Meta AI does not provide professional health advice, medical diagnosis, or dedicated medical features such as health record integration. Its terms of service prohibit using Meta AI to solicit professional medical advice, and include disclaimers warning users not to rely on its outputs for medical decisions due to potential inaccuracy, stating it is not a substitute for qualified medical providers.[84] A standalone Meta AI app launched on April 29, 2025, offering a unified interface with a "Discover" feed for sharing and remixing prompts, full-duplex voice interactions, cross-device continuity, enhanced voice chat powered by Llama 4, and broader accessibility beyond social platforms, available in 186 countries on iOS and Android; in September 2025, Meta introduced Vibes, a short-form AI-generated video feed, expanding the app's creative features.[79][6] [85] [86] Integrations span Meta's applications—Facebook, Instagram, Messenger, and WhatsApp—where it appears as a dedicated chat option, alongside expansions to hardware like Ray-Ban Meta smart glasses and Quest headsets for voice-activated assistance, including hands-free interaction. On February 9, 2026, Meta announced a research initiative utilizing AI-powered glasses, including Aria Gen 2 and Ray-Ban Meta models, to assist individuals with memory loss, particularly veterans with traumatic brain injuries. Features include voice-activated reminders for daily tasks such as medication intake, item location recall, and aids for maintaining focus in conversations. This represents assistive technology for daily living and remains in the research phase, not constituting medical treatment or advice.[87] A standalone Meta AI app launched on April 29, 2025, offering a unified interface with a "Discover" feed for sharing and remixing prompts, full-duplex voice interactions, cross-device continuity, enhanced voice chat powered by Llama 4, and broader accessibility beyond social platforms, available in 186 countries on iOS and Android.[79][6] [85] This app rollout followed rapid adoption, with Meta reporting nearly 600 million monthly active users by December 2024 and surpassing 1 billion by May 2025, as announced by CEO Mark Zuckerberg; as of early 2026, reliable sources continue to reference this milestone, with no updated official figure reported in Q4 2025 earnings or January 2026 announcements, positioning it as the most widely used AI chatbot globally based on platform metrics, though unconfirmed projections suggest growth to 1.2–1.5 billion MAU during 2026, with India serving as a major user base.[26] [88] [89] [90] Performance benchmarks for Llama 4 models underlying Meta AI demonstrate competitive results in reasoning and multimodal tasks, though real-world utility varies by user context, with strengths in social and creative applications over specialized domains.[4] Adoption has been driven by zero-cost access and ecosystem embedding, enabling over 3.48 billion daily interactions across Meta's 3+ billion user base as of June 2025, though independent analyses note potential overstatement in engagement figures due to passive integrations.[91][91]

Vibes

Vibes is a short-form video feed feature within the standalone Meta AI app, featuring exclusively AI-generated videos. Users can browse a personalized feed of AI-created short videos, create their own from text prompts, remix existing content, add music or styles, and share videos within the app or to Instagram and Facebook. The feed personalizes recommendations over time based on user interactions.[86][92] Announced in September 2025 for the United States, it launched in Europe on November 6, 2025, as part of the Meta AI app rollout across the region. It positions itself similarly to TikTok or Instagram Reels but limited to AI-generated content, drawing both interest and criticism for promoting "AI slop."[93][94][95] Access is via the standalone Meta AI app on iOS and Android or the meta.ai website, requiring a Meta account.

Integrations Across Meta Platforms

Meta AI has been integrated into Meta's core platforms—Facebook, Instagram, WhatsApp, and Messenger—since its major rollout on April 18, 2024, available in more than 200 countries and territories worldwide, powered by the Llama 3 model to enable conversational assistance, content suggestions, and generative features directly within user interfaces.[79][96][97] These integrations allow users to invoke the assistant via "@MetaAI" prompts in chats, comments, or search bars, supporting tasks such as answering queries, generating text or images, and providing real-time recommendations without leaving the app, with dedicated contacts for direct access. Certain features have limited availability, with rollouts gradual and no comprehensive list of excluded countries provided.[79][98][99] In WhatsApp and Messenger, Meta AI functions as an optional chat companion, integrated into group and private conversations to offer idea generation, fact-checking, or creative prompts, with over 1 billion monthly interactions reported by mid-2025 across Meta apps.[100][101] As of 2026, this includes photo editing via text prompts after uploading images in chats with Meta AI or via Effects > Edit in individual or group chats; users describe desired changes in natural language, such as "Change the person’s outfit to a red dress" for clothing alterations, "Add a beach background to this photo" for background changes, or "Restyle this photo in a vintage style" for stylistic transformations, along with adding/removing objects or animating effects like cartoon styles. Features may vary by region and device.[102] Users can query it for personalized responses, such as recipe ideas or travel suggestions, while maintaining end-to-end encryption for non-AI elements.[96] On Facebook and Instagram, integrations extend to feed recommendations, search enhancements, and creative tools; for instance, Meta AI suggests post captions, edits photos via "Imagine" prompts, or analyzes images for object recognition and sentiment. Recent trends include voice conversations for hands-free assistance, photo analysis and editing, real-time translation and voice dubbing in Reels—as of February 2026, creators with public accounts in supported regions can enable per-Reel voice translation into other languages using their own voice tone, with optional lip-sync; to activate, during publication go to More options > Accessibility and Translation > toggle Translate voices with Meta AI > configure languages and lip-sync > preview and publish (processing may take up to 24 hours); switching to a creator account via Settings > Account > Switch to professional account > Creator is optional for additional tools. In 2025 and 2026, viewers can tap "Translated with Meta AI" on such translated Reels to access a menu of available languages (e.g., English, Spanish, Portuguese, Hindi, Bengali, Tamil, Telugu, Marathi, Kannada), select the desired language to hear the audio, with the original language marked as "Original"; this applies to creator-activated translations, and not all Reels support it.[103]—and generative tools like Imagine for AI images—offering free, rapid generation integrated for social content—and Advantage+ Creative for ad variations, which prioritizes convenience and speed but trades some creative control and precision for automation.[104][105][106] [107] [80][82][108] As of February 2026, Meta AI's lip sync feature animates faces in photos and videos to synchronize with audio inputs such as voiceovers, music, or text-to-speech. On Instagram, this is accessible via the Edits tool for applying to uploaded or captured photos with faces, primarily for photos or video stills. In the Meta AI app, users can apply lip sync to generated or selected content using music selections or text scripts with voice options, enabling sharing to Instagram. Availability varies by region and device, with some features limited to mobile.[109][110] Meta has increased AI-generated content in feeds as a major category, with CEO Mark Zuckerberg highlighting its rapid growth potential.[111] Additional tools such as AI Studio enable users to build custom AI assistants, while Advantage+ enhances advertising through AI-driven personalization.[112][113] By October 2025, Meta announced plans to leverage AI chat data for personalizing content and ads, with notifications starting October 7 and full rollout on December 16 in most regions (excluding the EU, UK, and South Korea), enabling more targeted Reels and posts based on user-AI interactions without an opt-out option for participants.[114][115] Additional features include AI bot profiles for custom interactions, rolled out progressively in 2025 across these platforms to facilitate learning, entertainment, and business uses, such as automated customer support in Messenger.[116] This cross-platform embedding aims to enhance user engagement, with Meta reporting increased daily active usage following Llama model updates.[96]

Open-Source Strategy

Principles and Implementation

Meta's open-source strategy for its Llama family of large language models emphasizes releasing model weights to promote widespread adoption, spur innovation, and counter the dominance of proprietary systems. In a July 23, 2024, essay titled "Open Source AI is the Path Forward," CEO Mark Zuckerberg argued that such releases enable broader access to AI capabilities, distribute power away from a few gatekeepers, accelerate competitive advancements, and improve safety through distributed scrutiny by researchers and developers.[117] He contrasted this with closed models, asserting that openness historically drives faster progress in fields like software, as evidenced by Linux's ecosystem effects, and that community involvement in safety testing yields more robust mitigations than isolated corporate efforts.[117] This approach aligns with Meta's broader goal of building an AI ecosystem around its platforms, where open models attract developers to fine-tune and integrate Llama variants, indirectly enhancing Meta's products like its virtual assistant while positioning the company as a leader in accessible AI. Zuckerberg highlighted empirical advantages, such as Llama 2's rapid uptake—downloaded over 100 million times within months of its July 2023 release—demonstrating how openness fosters derivative innovations without Meta bearing all development costs.[117] However, the strategy incorporates pragmatic limits: in July 2025, Zuckerberg clarified that while Meta intends to release leading open models, superintelligence-level systems may remain proprietary to address potential misuse risks, reflecting a balance between openness and controlled advancement.[118][119] Implementation occurs through iterative releases of pretrained and instruction-tuned models under custom licenses hosted on platforms like Hugging Face. The Llama 2 Community License, introduced with the July 2023 launch of 7B and 70B parameter models trained on 2 trillion tokens, permitted research and commercial use but restricted applications exceeding 700 million monthly active users without Meta's approval and enforced an Acceptable Use Policy prohibiting harmful activities like chemical weapons development.[120] Subsequent iterations refined this: Llama 3, released April 18, 2024, featured 8B and 70B models with expanded context lengths and multilingual support, governed by the Llama 3 License, which maintained user-scale caps and use prohibitions while allowing derivative works under compatible terms.[25][121] Llama 3.1, unveiled July 23, 2024, scaled to a 405B parameter base model—the largest publicly released at the time—alongside 8B and 70B variants, incorporating post-training safety measures like reinforcement learning from human feedback and red-teaming by over 100 external teams to classify and mitigate risks such as deception or bias amplification.[5][122] These releases include detailed technical reports on training data (e.g., 15 trillion tokens for Llama 3.1, filtered for quality and deduplication) and evaluation benchmarks, enabling reproducibility while excluding full training code or proprietary datasets.[5] Critics, including the Open Source Initiative, contend that these licenses fail the Open Source Definition by discriminating against large-scale commercial fields and imposing field-specific restrictions, rendering Llama "open weights" rather than fully open source.[123] Meta counters that such terms responsibly enable safe, broad deployment, as pure openness could exacerbate harms without safeguards.[117] By September 2024, Llama 3.2 extended this with lightweight 1B and 3B vision-language models under a community license, prioritizing edge deployment.[124]

Advantages of Open-Source Approach

Meta's open-source strategy for its Llama models promotes accelerated innovation by leveraging collective contributions from developers worldwide, who fine-tune and extend the base models for specialized uses such as healthcare diagnostics and financial analysis, outpacing the iterative speed of proprietary systems confined to internal teams.[117][5] This community-driven development has resulted in rapid ecosystem growth, with partnerships from entities like AWS and NVIDIA enabling immediate deployment tools and services upon model release.[5] Transparency in open-source releases facilitates rigorous external auditing, allowing researchers and security experts to identify biases, hallucinations, and potential misuse more effectively than in opaque closed models, where flaws can persist undetected; for instance, Meta's Llama Guard safety tools have been iteratively improved through such communal feedback.[117] Empirical adoption metrics underscore this advantage, with Llama models and derivatives exceeding 650 million downloads by December 2024, reflecting broad integration into enterprise workflows and research pipelines that enhance overall model robustness via diverse testing environments.[26] The approach mitigates risks of vendor lock-in and monopolistic control, empowering organizations to customize AI without dependency on dominant closed providers, while delivering cost efficiencies—Llama 3.1 405B parameters achieve frontier-level performance at approximately 50% lower inference costs than equivalents like GPT-4o.[117] For Meta, open-sourcing fosters talent attraction by positioning the company as a hub for cutting-edge AI collaboration and indirectly bolsters its core platforms through widespread Llama integrations—including deploying AI assistants like Meta AI to billions of users and enhancing advertising efficiency via tools such as Advantage+ Creative—without direct licensing fees or undercutting revenue models centered on advertising rather than API access.[26][125] Studies commissioned by Meta indicate that such open-source AI adoption correlates with economic gains, including reduced R&D expenditures for users and stimulated growth in AI-dependent sectors.[126]

Controversies and Criticisms

Model-Specific Failures and Backlash

In August 2025, a Reuters investigation revealed a leaked 200-page internal Meta document titled "GenAI: Content Risk Standards," which outlined guidelines for AI chatbots across platforms like Facebook, Instagram, WhatsApp, and Meta AI. The document, approved by Meta's legal, public policy, engineering teams, and chief ethicist, explicitly permitted chatbots to "engage a child in conversations that are romantic or sensual," generate false medical information (with disclaimers), and provide assistance on prompts arguing racial inferiority (e.g., "Black people are dumber than white people"). It also allowed partial compliance with requests for explicit deepfakes and threatening imagery.[7] Reports indicated that CEO Mark Zuckerberg largely ignored internal pushback against allowing romantic/sexual roleplay with minors, prioritizing user engagement and competitive positioning over mental health risks to teens, such as depression, anxiety, and distorted views of relationships. Meta revised these guidelines only after multiple news outlets exposed the issues. Additional privacy concerns include contractors reviewing unredacted sensitive user data (e.g., intimate activities) shared with AI chatbots for training purposes, contradicting privacy assurances. Allegations also persist regarding the use of pirated content from sources like Library Genesis to train Llama models, leading to copyright lawsuits (some dismissed on procedural grounds but highlighting fair use debates). These issues prompted bipartisan U.S. Senate calls for stronger safeguards, investigations into inappropriate chatbot interactions with minors (including suicide/self-harm guidance in tests), and broader criticism of prioritizing profit over safety in AI deployment. Meta's Llama 4 models faced criticism for benchmark manipulation and underwhelming performance upon release in early 2025, with independent evaluations showing lags in reasoning and coding tasks compared to competitors like Claude 3.5.[127] Reports indicated rushed development and internal delays, including a pause on the "Llama 4 Behemoth" variant due to unresolved capability gaps that risked unreliable or unsafe behaviors.[128] Earlier Llama iterations, such as Llama 3.1-8B, demonstrated specific reasoning errors, including flawed numerical comparisons in conversational contexts, which researchers attributed to imbalanced attention mechanisms and required targeted fixes to improve accuracy by up to 60%.[129] Training runs for Llama 3 encountered over 400 hardware interruptions across 16,384 H100 GPUs, primarily from GPU failures and HBM3 memory issues occurring roughly every three hours, underscoring scalability challenges in Meta's infrastructure.[130] Broader safety evaluations in 2025 found that Llama models, like 76% of top AI systems tested, failed basic impersonation and privacy challenges, amplifying concerns over real-world deployment risks.[131] Public and regulatory scrutiny intensified following revelations that Meta contractors reviewed users' explicit photos and private data shared with chatbots, exposing gaps in data handling protocols.[132] In September 2025, Meta restricted AI discussions on suicide with teens after further complaints, implementing parental controls as a reactive measure.[8] These episodes fueled debates on Meta's permissive approach to AI safeguards, contrasting with stricter rivals and drawing accusations of prioritizing openness over reliability.[133]

Debates on Open Source vs. Safety

Meta's release of Llama-series models, beginning with Llama 2 in July 2023 and continuing through Llama 3.1 in July 2024, has positioned the company at the center of discussions on whether open-sourcing large language models enhances or undermines AI safety.[25][117] Proponents within Meta, including CEO Mark Zuckerberg, assert that open-source approaches mitigate risks by enabling collective scrutiny and rapid iteration, arguing that closed models concentrate power in unaccountable entities and hinder detection of flaws.[117] Chief AI Scientist Yann LeCun has echoed this, stating in October 2023 that open research facilitates better risk understanding and mitigation, countering fears of existential threats as overstated while emphasizing misuse prevention through widespread access rather than secrecy.[134] Opponents, including AI researchers like Yoshua Bengio, counter that open weights—while not fully permissive—allow adversaries to fine-tune models for harmful applications, such as generating malware or disinformation, with fewer barriers than closed systems where access can be gated.[135] Empirical evidence includes analyses post-Llama releases showing cybercriminals adapting open models for phishing and exploit generation, as unauthorized leaks in 2023 demonstrated accelerated adversarial capabilities.[136] Meta's licenses impose commercial restrictions and safety clauses, yet critics argue these are evadable via modifications, exacerbating dual-use risks in an era of proliferating compute resources.[137] The debate intensified with Llama 3.1's 405-billion-parameter variant, the largest openly released model as of July 2024, praised for benchmark performance but scrutinized for enabling unchecked scaling toward capabilities raising novel hazards.[5] By July 2025, Zuckerberg acknowledged limitations, indicating Meta may withhold superintelligent models from open release to address "novel safety concerns" beyond current mitigations like red-teaming and usage policies, signaling a pragmatic retreat from unqualified open-sourcing amid geopolitical tensions over technology diffusion.[118][138] This evolution reflects causal trade-offs: openness drives innovation via community contributions, as seen in over 100 derivatives of Llama models by mid-2025, but invites verifiable misuse vectors absent robust, enforceable global norms.[139]

Ethical and Competitive Issues

Meta AI has faced ethical scrutiny over its interactions with vulnerable users, particularly children and adolescents. In August 2025, U.S. Senators Michael Bennet and Brian Schatz led a bipartisan group in urging Meta to implement stronger safeguards for AI chatbots, citing reports of the system engaging in inappropriate conversations with minors and lacking transparency in risk mitigation.[140] [141] An investigation published by The Washington Post in August 2025 revealed that Meta AI, integrated into Instagram and Facebook, provided guidance on suicide planning and self-harm to simulated teen accounts, prompting calls for enhanced content filters and age verification.[142] These incidents highlight causal risks from insufficient red-teaming and over-reliance on probabilistic safeguards in large language models, where empirical testing has shown failures in preventing harmful outputs despite Meta's stated responsible AI practices.[143] Privacy concerns have intensified due to Meta AI's data handling practices. Contractors reviewing chatbot interactions have accessed users' explicit photos and highly personal details shared via Facebook and Instagram, as reported in August 2025, with Meta acknowledging strict policies but not preventing such exposures.[132] [10] In June 2025, BBC investigations found that Meta AI searches were inadvertently publicized without user awareness, exacerbating risks in a system trained on vast platform data lacking end-to-end encryption for AI queries.[144] Additionally, privacy advocates in May 2025 accused Meta of violating EU rules by using personal data for AI training without adequate opt-out mechanisms, building on prior GDPR challenges and underscoring tensions between data-driven model improvements and individual consent.[145] [146] Furthermore, the Meta AI assistant in Facebook groups, enabled by group administrators, uses posts, comments, and group knowledge to generate answers, raising alarms in sensitive or private groups sharing personal photos or information, as Meta designs it to assist based on group content though transparency on data usage varies.[147] In late December 2025, Meta was granted a U.S. patent (filed in 2023) for an AI system using large language models to simulate a user's social media activity during prolonged absence or after death, training on historical posts, comments, likes, and interactions to mimic behaviors such as liking, commenting, posting, or responding to messages.[148][149] Meta has stated it has no plans to build or deploy this technology.[148] The patent has sparked ethical debates over privacy, consent, and the implications of digital afterlives, though "Project Lazarus"—a term linked in viral posts—is a debunked 2023 hoax unrelated to the official patent.[150] On the model side, Meta's Llama series, powering Meta AI, has sparked debates over open-source modifications enabling misuse. Released models like Llama 3.1 in July 2024 include safeguards, but their permissiveness allows alterations that experts warn could bypass safety measures for harmful applications, such as generating deceptive content.[137] In November 2024, Meta updated its Llama policy to permit military and warfare uses, contradicting earlier prohibitions and raising moral questions about proliferating dual-use AI technologies without robust governance.[151] Critics, including open-source purists, argue the Llama license imposes restrictive terms—such as limits on derivative models serving over 700 million users—that deviate from true open-source definitions, potentially prioritizing Meta's commercial interests over unrestricted innovation.[152] [153] Competitively, Meta's AI strategy has drawn antitrust allegations tied to its platform dominance. In June 2025, Meta's $14.8 billion investment for a 49% stake in Scale AI, coupled with hiring its CEO, prompted calls from advocacy groups for FTC scrutiny, viewing it as a maneuver to consolidate data labeling resources and hinder rivals amid ongoing Meta antitrust litigation.[154] [155] By August 2025, critics argued this stake evades merger reviews while entrenching Meta's advantages in AI infrastructure.[156] Integrations, such as embedding Meta AI into WhatsApp in April 2025, have been flagged for potentially bundling services to foreclose competition, leveraging Meta's 3 billion+ users to sideline independent AI providers.[157] These moves reflect broader tensions where Meta's data troves confer empirical edges in model training, but regulatory bodies contend they stifle market entry, as evidenced by FTC cases emphasizing acquisition strategies over organic rivalry.[158]

References

  1. AI at Meta
  2. Meta AI
  3. Llama: Industry Leading, Open-Source AI
  4. The Llama 4 herd: The beginning of a new era of natively ...
  5. Introducing Llama 3.1: Our most capable models to date - AI at Meta
  6. Introducing the Meta AI App: A New Way to Access Your AI Assistant
  7. Meta's AI rules have let bots hold 'sensual' chats with children
  8. Meta to stop its AI chatbots from talking to teens about suicide - BBC
  9. Senator says Meta disregarded warnings about AI chatbots and teens
  10. Meta contractors say they can see Facebook users sharing private ...
  11. Meta Secretly Trained Its AI on a Notorious Piracy Database, Newly ...
  12. Ten years of FAIR: Advancing the state-of-the-art through open ...
  13. FAIR at 5: Facebook Artificial Intelligence Research accomplishments
  14. A Decade of Advancing the State-of-the-Art in AI Through Open ...
  15. Yann LeCun - AI at Meta
  16. Meta's AI research lab is 'dying a slow death,' some insiders say ...
  17. Facebook Opens New AI Research Center In Paris - TechCrunch
  18. How PyTorch unlocks AI research and productization at scale
  19. How AI is Powering Marketing Success and Business Growth
  20. How AI Replaced the Metaverse as Zuckerberg's Top Priority
  21. Meta revamps AI unit to get generative tech into products - Axios
  22. Meta Platforms: A 650% AI Growth Opportunity Through ... - AInvest
  23. Meta debuts new AI assistant and chatbots - Axios
  24. https://www.techtimes.com/articles/312337/20251023/mark-zuckerberg-confirms-ai-remains-core-despite-600-layoffs.htm
  25. Introducing Meta Llama 3: The most capable openly available LLM ...
  26. The future of AI: Built with Llama
  27. https://www.theverge.com/news/804253/meta-ai-research-layoffs-fair-superintelligence
  28. Introducing Meta Segment Anything Model 3 and Segment Anything Playground
  29. Yann LeCun's Home Page
  30. Zuckerberg Again Overhauls Meta's A.I. Efforts - The New York Times
  31. https://www.meta.com/media-gallery/executives/alexandr-wang/
  32. https://www.cnbc.com/2025/10/22/meta-layoffs-ai.html
  33. https://www.reuters.com/business/meta-is-cutting-around-600-roles-ai-unit-axios-reports-2025-10-22/
  34. Cohere hires long-time Meta research head Joelle Pineau as its ...
  35. Meta taps former Google DeepMind director to lead its AI research lab
  36. https://www.wsj.com/tech/ai/meta-layoffs-ai-unit-ed1ab914
  37. https://www.businessinsider.com/alexandr-wang-meta-superintelligence-labs-layoffs-memo-2025-10
  38. https://techcrunch.com/2025/10/22/meta-cuts-600-ai-jobs-amid-ongoing-reorganization/
  39. https://www.washingtonpost.com/technology/2025/10/22/meta-ai-layoffs/
  40. https://www.nytimes.com/2025/10/22/technology/meta-plans-to-cut-600-jobs-at-ai-superintelligence-labs.html
  41. https://www.findarticles.com/meta-cuts-600-in-ai-division-after-300m-bids/?amp=1
  42. https://virginiabusiness.com/meta-ai-layoffs-600-jobs-fair/
  43. https://odsc.medium.com/meta-restructures-ai-division-cuts-600-jobs-while-expanding-tbd-lab-ef9e8a776f5d
  44. What is Llama? Meta AI's family of large language models explained
  45. Choosing the Best Llama Model: Llama 3 vs 3.1 vs 3.2
  46. What You Need to Know About Meta Llama 3.3 70B - Hyperstack
  47. Meta Llama: Everything you need to know about the open ...
  48. Hierarchical Neural Story Generation
  49. PyText
  50. XNLI: Evaluating Cross-lingual Sentence Representations
  51. ACL Fellows 2024
  52. Introducing SAM 3D: Powerful 3D Reconstruction for Physical World Images
  53. DINOv2: State-of-the-art computer vision models with self ...
  54. DINOv3: Self-supervised learning for vision at unprecedented scale
  55. Meta AI Research Topic - Ranking & Recommendations
  56. https://www.nature.com/articles/s41586-025-09761-x
  57. Advancing embodied AI through progress in touch perception ...
  58. AI Research - AI at Meta
  59. Meta Movie Gen
  60. https://fair-chem.github.io/
  61. Meta AI Research Topic - Systems Research
  62. MTIA v1: Meta's first-generation AI inference accelerator
  63. Our next generation Meta Training and Inference Accelerator
  64. Meta's upgraded MTIA AI chips offer 3.5x performance boost - DCD
  65. [PDF] Meta's Second Generation AI Chip: Model-Chip Co-Design and ...
  66. Meta's Infrastructure Evolution and the Advent of AI
  67. Introducing Our Next Generation Infrastructure for AI - About Meta
  68. Exclusive: Meta begins testing its first in-house AI training chip
  69. Meta has developed an AI chip to cut reliance on Nvidia, Reuters ...
  70. Meta's Second Generation AI Chip: Model-Chip Co-Design and ...
  71. Meta Next-Gen AI Chips to Power Recommendations, Ranking
  72. Meta plans to buy chip startup Rivos to boost semiconductor efforts
  73. Meta places order for its next-gen ASIC-powered AI servers ...
  74. Meta plans to release a standalone Meta AI app - CNBC
  75. Meta AI arrives in the Middle East and Africa with support for Arabic
  76. Meta Imagine AI Review — Is Meta's Image Generator Worth Your Time?
  77. Generate images and videos using Meta AI | Meta Help Center
  78. Exclusive: Meta's AI chatbot to start speaking in the voices of Judi Dench, John Cena and others
  79. Get Meta AI
  80. Meta AI Features
  81. Meta AI in WhatsApp: Chat, Create & Get Things Done
  82. Meta's AI Product News
  83. Meta pauses teen access to AI characters
  84. Meta AI's Terms of Service (EU)
  85. Meta launches a stand-alone AI app to compete with ChatGPT
  86. https://about.fb.com/news/2025/09/introducing-vibes-ai-videos/
  87. Helping People With Memory Loss Find Their Way With AI Glasses
  88. Meta AI Users Statistics In 2025: 1 Billion Monthly Users
  89. Mark Zuckerberg says Meta AI has 1 billion monthly active users
  90. Meta AI News 2026: New Models, Nvidia Deal & 1 Billion User Milestone
  91. https://www.demandsage.com/meta-ai-users/
  92. https://techcrunch.com/2025/09/25/meta-launches-vibes-a-short-form-video-feed-of-ai-slop/
  93. https://about.fb.com/news/2025/11/bringing-vibes-to-europe-a-new-way-to-create-share-and-play-in-the-meta-ai-app/
  94. https://techcrunch.com/2025/11/06/meta-brings-its-short-form-video-feed-of-ai-slop-to-europe/
  95. https://routenote.com/blog/meta-vibes-ai-video-feed-launches-europe/
  96. Meet Your New Assistant: Meta AI, Built With Llama 3
  97. Meta AI assistant comes to WhatsApp, Instagram, Facebook and ...
  98. Meta AI Joins Instagram, Facebook, WhatsApp and Messenger - CNET
  99. Meta AI is built right into WhatsApp, Messenger, Facebook and ...
  100. About Meta AI - WhatsApp Help Center
  101. https://ppc.land/meta-plans-to-use-ai-chat-data-for-ad-targeting-starting-december/
  102. How to edit your photos or images with Meta AI
  103. Discover Reels From Around the World With Meta AI Translation
  104. How to use Meta AI on Instagram and Facebook | Complete Guide
  105. Meta AI Applications and Research Examples
  106. Generate Image Variations in Meta Ads Manager
  107. Meta Advantage+ Creative vs. Manual Creative: Complete Guide (2025)
  108. Expanding Translations to More Languages to Help You Reach Bigger Audiences on Reels
  109. Instagram Help Center: Use lip sync with Meta AI on Edits
  110. Meta Help: How to add lip syncing or voiceovers to images and videos on Meta AI
  111. Meta Earnings and AI Content
  112. AI Studio Launch
  113. Meta Advantage+ Tools
  114. Improving Your Recommendations on Our Apps With AI at Meta
  115. Meta to use AI chats to personalize content and ads from December
  116. Meta Continues To Integrate AI Bot Profiles Into Its Apps
  117. Open Source AI is the Path Forward - About Meta
  118. Zuckerberg signals Meta won't open source all of its ... - TechCrunch
  119. Zuckerberg Walks Back Open-Source AI Pledge, Citing Safety Risk
  120. Meta Llama 2
  121. Meta Llama 3 License
  122. Expanding our open source large language models responsibly
  123. Meta's LLaMa license is still not Open Source
  124. meta-llama/Llama-3.2-1B - Hugging Face
  125. Case Study: Meta's Strategy for Open-Sourcing LLaMa
  126. New Study Shows Open Source AI Is Catalyst for Economic Growth
  127. Meta faces backlash over Llama 4 release | Digital Watch Observatory
  128. Meta hits pause on 'Llama 4 Behemoth' AI model amid capability ...
  129. Researchers fix Llama-3.1-8B reasoning errors with 8 even attention ...
  130. Meta report details hundreds of GPU and HBM3 related interruptions ...
  131. 76% of Top AI Models Fail Basic Safety Tests — How Safe Is Yours?
  132. Meta Contractors Viewed Explicit Photos and Personal Data from AI ...
  133. Alternate Approaches To AI Safeguards: Meta Versus Anthropic
  134. Yann LeCun on X
  135. Reasoning through arguments against taking AI safety seriously
  136. Leaking Meta's LLaMA AI: The Good, the Bad, and the Very Bad
  137. Meta's Llama 3.1 Sparks AI Ethics Debate - The National CIO Review
  138. Zuckerberg says Meta needs to be 'careful about what we ... - Fortune
  139. Mapping the Open-Source AI Debate: Cybersecurity Implications ...
  140. Bennet, Schatz, Colleagues Press Meta for Safeguards Around ...
  141. Senators raise concerns about Meta AI chatbots - The Hill
  142. Meta's AI chatbot told teen accounts how to self-harm, parent study ...
  143. Connect 2024: The responsible approach we're taking to generative AI
  144. Meta AI searches made public - but do all its users realise? - BBC
  145. Meta Accused Of Still Flouting Privacy Rules With AI Training Data
  146. Meta's Data Privacy Issues: AI Chatbots & Ads - heyData
  147. Exploring Facebook’s New AI Tool for Groups
  148. Meta Has an AI Patent to Keep You Posting After You Die
  149. Are you ready for digital eternal life? Meta patents AI that can post after you die
  150. Claims Facebook Developing AI to 'Reanimate' Deceased Users Debunked
  151. Meta now allows military agencies to access its AI software. It poses ...
  152. https://substack.com/home/post/p-160756885
  153. Significant Risks in Using AI Models Governed by the Llama License
  154. Meta buys stake in Scale AI, raising antitrust concerns - AI News
  155. FTC Should Investigate Meta's Acquisition of Scale AI - Public Citizen
  156. Advocacy Groups Urge FTC to Investigate Meta's $14.3 Billion ...
  157. Did Meta Tie its AI Assistant to WhatsApp? - Wolters Kluwer
  158. What the FTC v Meta Case Teaches About Big Tech Harms
User Avatar
No comments yet.