As the global push to advance artificial intelligence continues to gain momentum, Meta has officially set its sights on one of the field’s most ambitious goals: Superintelligence, AI that surpasses human intelligence in every way imaginable. To get there, the company has consolidated its AI projects into a new division called Meta Superintelligence Labs (MSL), with the ultimate goal of creating what CEO Mark Zuckerberg calls “personal superintelligence,” where AI is built for individual fulfillment rather than just enterprise or research applications.
Meta Superintelligence Labs, Explained
Meta Superintelligence Labs (MSL) is Meta’s new division that houses all its AI teams and initiatives, including the development of its Llama language models and its fundamental and applied AI research. Together, these teams will work toward the realization of superintelligence — AI that exceeds human intelligence in all aspects.
The move comes on the heels of a disappointing showing from Meta’s Llama 4 models, which left the company trailing behind rivals like OpenAI, Anthropic, Google and DeepSeek. To close the gap, Zuckerberg has gone on an aggressive hiring spree, offering signing bonuses of up to $1 billion to lure some of the industry’s top players over to MSL. One of the most pivotal hires so far is former Scale AI CEO Alexandr Wang, who was brought on as part of Meta’s $15 billion acquisition of Scale AI.
While overhauling Meta’s approach to AI has come at a high cost, it reflects an increasingly heated competition for AI talent and resources that Zuckerberg is determined not to lose. If this gamble pays off, the Superintelligence Labs could revitalize the company’s AI ambitions and propel it back to the forefront of the industry.
Here’s what we know so far about Meta’s Superintelligence Labs and what it could mean for the future of AI development.
What Is Meta Superintelligence Labs?
Meta Superintelligence Labs is a new company division that houses teams working on Meta’s foundation models, AI research projects and AI products. It will be made up of four teams — known as TBD Lab, FAIR, Products and Applied Research and MSL Infra — who will coordinate their efforts in the pursuit of superintelligence, a theoretical form of AI that exceeds human intelligence in every aspect, including creative thinking, problem-solving, decision-making and emotional intelligence.
Specifically, MSL is focused on the development of what Mark Zuckerberg calls “personal superintelligence,” where the system knows its user “deeply,” understands their goals and helps to achieve them.
“In the coming years, AI will improve our existing systems and enable the creation and discovery of new things that aren’t imaginable today,” Zuckerberg wrote in a 2025 manifesto. “As profound as the abundance produced by AI may be one day, an even more meaningful impact on our lives will likely come from everyone having a personal superintelligence that helps you achieve your goals, create what you want to see in the world, experience any adventure, be a better friend to those you care about, and grow to become the person you aspire to be.”
Zuckerberg frames this as almost a necessity, especially given that artificial intelligence is on track to completely transform the job market. AI-related workforce reductions are already underway, and the trend is only going to accelerate as automation encroaches on more kinds of work — specifically white collar desk jobs, and even more specifically at the entry level. Everyday workers are panicking, but Zuckerberg and other AI evangelists seem confident that this will ultimately be a good thing.
“If trends continue, then you’d expect people to spend less time in productivity software, and more time creating and connecting,” Zuckerberg continued. “Meta’s vision is to bring personal superintelligence to everyone. We believe in putting this power in people’s hands to direct it towards what they value in their own lives.”
What We Know About Meta Superintelligence Labs So Far
Exactly how Meta will fulfill Zuckerberg’s vision for personal superintelligence remains to be seen, but details are beginning to emerge with the company’s current projects and Zuckerberg’s own statements.
As of August 19, 2025, Meta Superintelligence Labs restructured into four teams to accelerate the company’s pursuit toward superintelligence. These teams include:
- TBD Lab, a research unit led by Meta’s chief AI officer Alexandr Wang tasked with developing the Llama language models that power the Meta AI assistant.
- FAIR (short for fundamental AI research), a long-standing internal AI research team led by Meta’s Director of AI Research Rob Fergus that focuses on long-term projects and achieving “advanced machine intelligence.”
- Products and Applied Research, a team led by former GitHub CEO Nat Friedman that will work to integrate Llama models and AI research into Meta consumer products.
- MSL Infra, a team led by Meta’s vice president of engineering Aparna Ramani that will focus on the AI infrastructure needed to support Meta’s AI goals.
MSL will also continue Meta’s fundamental and applied AI initiatives, including research on its so-called “world models” — AI models that understand the physical dynamics of the real world and predict how interactions between different objects will play out. So far, this has culminated in Video Joint Embedding Predictive Architecture 2 (V-JEPA 2), which Meta released in June 2025, boasting its advanced visual learning and predictive capabilities.
Zuckerberg also revealed that Meta has witnessed “early glimpses of self-improvement with the models.” To give its models the necessary fuel to continue progressing, the company is wasting no time in building out its AI infrastructure. It has even reportedly been using tents to speed up the construction of its new Ohio-based data center, Prometheus, which is slated to be up and running by 2026.
But there’s a caveat to this rapid push for superintelligence: In his manifesto, Zuckerberg stated the company needs to be “careful about what we choose to open source,” citing “novel safety concerns.” The prospect of Meta moving away from an open-source approach to AI only raises more concerns about the company’s direction and the dangers of pursuing superintelligence in general.
How Does Meta Superintelligence Labs Differ From Other AI Projects?
Meta’s explicit pursuit of superintelligence is unprecedented in that it is bringing the concept into the corporate mainstream. Many companies like Google DeepMind and xAI are still working to achieve artificial general intelligence (AGI), a tier below superintelligence that can mimic the way humans learn and perform tasks it was never trained to do. And while Sam Altman has tried to shift OpenAI’s attention to superintelligence, the company’s contract with Microsoft could drag it into a legal conflict and hamper Altman’s ambitions.
Perhaps the most defining trait that sets this project apart from others, though, is Meta’s goal of using superintelligence to promote “personal empowerment.” Rather than trying to automate business processes like most other AI companies, Zuckerberg emphasizes the idea of providing each individual with access to superintelligence to reach their own goals.
“This is distinct from others in the industry who believe superintelligence should be directed centrally towards automating all valuable work, and then humanity will live on a dole of its output,” Zuckerberg wrote in his manifesto. “At Meta, we believe that people pursuing their individual aspirations is how we have always made progress expanding prosperity, science, health, and culture.”
Criticisms of Meta Superintelligence Labs
Meta’s move to create its Superintelligence Labs has drawn plenty of skepticism, with shady work practices and excessive spending among critics’ main concerns.
Questionable Demands Made of Workers
Meta’s several multi-million-dollar salary offers to join MSL may have attracted top talent from rival companies, but they likely come with pretty major strings attached. Workers have reportedly had to sacrifice their work-life balance to help Meta catch up with leading AI competitors, and the company’s salary packages could depend on performance metrics that put even more pressure on employees.
Transition From Open-Source Models
While Zuckerberg cites safety concerns as the reason Meta might move away from open source, such a shift would likely cut off free public access to its models and concentrate even greater control over this technology in the hands of a few corporations. Limiting access would also put Meta in a more favorable position to profit from its products, prompting questions about whether this pivot is driven more by greed than caution.
After all, this would mark a notable change of heart for Meta, which has garnered a reputation as one of the industry’s most vocal champions of open-source AI. Zuckerberg even called it “the path forward” and “necessary for a positive AI future” in a 2024 blog post, arguing that transparency fosters innovation and safety. This stance helped distinguish Meta from rivals like OpenAI and Google, which (for the most part) make users pay for access to their most advanced models.
Doubts About Meta’s Investment Decisions
Outspending competitors doesn’t guarantee success. Meta’s metaverse efforts reveal as much, with the company’s Reality Labs reporting $4.53 billion in operating losses compared to only $370 million in sales during the second quarter of 2025. It remains to be seen whether the heavy investment in Superintelligence Labs will meet the same fate.
Elon Musk has even gone so far as to criticize Zuckerberg’s massive salary packages, suggesting that they’re unsustainable and unnecessary.
“Many strong Meta engineers have and are joining xAI and without the need for insane initial comp (still great, but not unsustainably high),” Musk wrote in a post on X. “And we are hyper merit-based: do something great and your comp can shift substantially higher.”
Potential Risks of Superintelligence
Zuckerberg’s mention of “novel safety concerns” only scratches the surface of just how powerful and unpredictable superintelligence could be. Today’s narrow intelligence already amplifies societal biases, accelerates the spread of misinformation and takes employment opportunities away from humans. Superintelligence could significantly magnify those problems — and create entirely new ones we haven’t yet imagined. There’s no telling what kind of impact it could have on the job market or what threats it could pose if it falls into the wrong hands, throwing into question the ethical implications of reaching superintelligence.
What Meta Superintelligence Labs Means for the Future of AI
The creation of Meta Superintelligence Labs ramps up the pressure on all AI companies to reach AGI and superintelligence. As part of this process, competitors may follow Meta’s playbook, hoarding AI talent and throwing around billions of dollars. And if Zuckerberg ends Meta’s open-source AI approach, this could encourage competitors to do the same, cutting out smaller companies, researchers and individuals while consolidating resources in the hands of a few major tech companies.
Given President Trump’s pro-AI stance, the industry could very well head in a more centralized direction. Among other things, his AI Action Plan calls for loosening federal regulations around AI and expanding the number of AI data centers in the United States — moves that stand to benefit tech titans like Meta. As a result, Superintelligence Labs could end up being part of a larger trend in which the biggest AI players are incentivized to make blunt power grabs in the race for AI supremacy.
Key Meta Superintelligence Labs Developments
Meta Superintelligence Labs has undergone several major changes since its formation, reflecting the company’s evolving approach to organizing AI research and advancing long-term goals toward artificial general intelligence.
Here’s some of MSL’s key developments to date:
Restructuring of Meta Superintelligence Labs (August 2025)
On August 19, Meta internally announced a major reorganization of its AI teams, dividing its Meta Superintelligence Labs into four separate teams: TBD Lab, FAIR, Products and Applied Research and MSL Infra. With this announcement, Meta’s AGI foundations team was also dissolved to focus on MSL initiatives.
The restructuring reflects CEO Mark Zuckerberg’s push to streamline Meta’s efforts in pursuit of “superintelligence,” a term he uses to describe the company’s next AI frontier.
The next day, the Wall Street Journal reported Meta froze AI hiring after the summer’s spending spree.
Creation of Meta Superintelligence Labs (June 2025)
In an internal memo, CEO Mark Zuckerberg announced the creation of Meta Superintelligence Labs, a division dedicated to developing artificial general intelligence, or “superintelligence” as it’s referred to by Meta.
MSL brought together teams from across Meta’s AI research and product groups under one umbrella — and recruited former talent from OpenAI and Google — with the goal of building AI systems that could eventually surpass human intelligence.
Zuckerberg framed the MSL initiative as the company’s most ambitious AI effort yet, emphasizing its long-term focus on scaling models and advancing reasoning capabilities.
Frequently Asked Questions
What is Meta superintelligence?
Meta Superintelligence Labs is a new AI division formed by Meta to coordinate its AI efforts toward achieving superintelligence, or advanced AI superior to human intelligence in every way possible. MSL will spearhead progress on Meta’s Llama language models and continue the company’s fundamental and applied AI research projects.
Who are the key hires in Meta Superintelligence Lab?
In addition to snatching up employees from rivals like OpenAI and Anthropic, Meta has also grown its headcount by acquiring startups like Scale AI and WaveForms. Here are some of the most prominent hires to join Meta Superintelligence Labs:
- Alexandr Wang, former CEO of Scale AI.
- Nat Friedman, former CEO of GitHub.
- Daniel Gross, former CEO of Safe Superintelligence.
- Joel Pobar, former technical staff member who worked on AI inference at Anthropic.
- Shuchao Bi, co-creator of GPT-4o mini and GPT-4o voice mode at OpenAI.
- Jack Rae, former Principal Research Scientist who worked on Gemini 2.5 at Google.
- Shengjia Zhao, former OpenAI researcher who contributed to several of the company’s biggest breakthroughs, including ChatGPT, GPT-4o and o1, its first reasoning model.
- Jason Wei and Hyung Won Chung, former OpenAI researchers who helped make the o3 and o1 reasoning models.
Who is in charge of Meta’s AI strategy?
Meta CEO Mark Zuckerberg is in charge of establishing the direction of the company, overseeing its overall AI strategy. New Chief AI Officers Alexandr Wang and Nat Friedman will take charge of the company’s Superintelligence Labs division, while Chief AI Scientist Yann LeCun continues leading Meta’s Facebook AI research initiatives.
How is Meta’s approach different from OpenAI or DeepMind?
OpenAI and Google DeepMind have yet to explicitly pivot to superintelligence, as both are still focused on reaching AGI. But the main distinction is that most AI companies are focused on using advanced AI to automate all types of work. In contrast, Meta’s mission revolves around “personal empowerment” and the idea of letting individuals use superintelligence to fulfill their own goals.
Is Meta planning to release its AI models to the public?
In his “personal superintelligence” manifesto, Mark Zuckerberg suggests that Meta may limit access to any superintelligence models it develops due to “novel safety concerns.” However, it remains to be seen how the company will approach its superintelligence models — or if they can even be achieved.
