Yann LeCun, one of Meta’s best-known AI leaders, is reportedly preparing to leave the company and build his own startup. The Financial Times reported the plan, citing anonymous sources, and said LeCun is already in talks to raise capital.
The reported move matters because it would come as Meta is reshaping its AI organization, shifting resources, and responding to concerns that rivals are moving faster. It also highlights a deeper question inside the AI industry: whether the next leap comes from today’s large language models, or from different approaches such as world models.
What the report says
LeCun is a professor at New York University, a senior researcher at Meta, and a winner of the prestigious A.M. Turing Award. At Meta, he is a chief AI scientist and has long been tied to the company’s Fundamental AI Research division, known as FAIR.
According to the Financial Times report, LeCun plans to leave in the coming months. The startup he is discussing with potential backers would continue his work on world models.
Meta did not immediately return a request for comment outside regular business hours. That means the central facts remain framed as reported plans, not a confirmed company announcement.
Why world models are central to the story
A world model is an AI system that builds an internal understanding of its environment. With that understanding, it can simulate cause and effect and use those simulations to predict outcomes.
That makes world models different from the way many people currently encounter AI through LLMs. The source article does not describe the full technical architecture of LeCun’s work, but it does make clear that world models are an active area of competition.
Top labs and startups are also developing world models, including Google DeepMind and World Labs. If LeCun builds a company around the idea, it would place his long-term research direction into a startup setting rather than inside Meta’s research structure.
For readers tracking AI development, the key point is simple: this is not only a personnel story. It is also about which AI research paths get capital, talent, and organizational focus.
Meta is already changing its AI structure
The reported departure would arrive during a pivotal period for Meta. The company has changed how it approaches AI development after concerns that it is being outpaced by OpenAI, Google, and Anthropic.
Meta has reportedly been revamping its AI organization after hiring over 50 engineers and researchers from competitors. Those hires are part of a new AI unit called Meta Superintelligence Labs, or MSL.
One major piece of that shift came in June, when Meta invested $14.3 billion in data-labeling vendor Scale AI. Meta also brought on Scale AI CEO Alexandr Wang to run the new division.
Those moves have reportedly created strain inside the company’s AI operation. Sources told TechCrunch in August that the unit had become increasingly chaotic, with new talent frustrated by the bureaucracy of a large company. At the same time, Meta’s previous generative AI team has seen its scope limited.
FAIR, MSL, and the long-term research question
LeCun’s work at Meta sits under FAIR, which is designed to focus on long-term AI research. The source describes that as research into techniques that may be used five to 10 years from now.
That mission differs from the newer MSL effort, which is part of Meta’s broader push after its previous family of AI models, Llama 4, failed to keep up with rival models. CEO Mark Zuckerberg’s decisions to overhaul the company’s AI approach have overshadowed LeCun’s long-term research work, according to the source article.
This creates a clear tension. FAIR represents patient research aimed at future techniques. MSL represents a more urgent organizational response to competitive pressure in the current AI race.
If LeCun leaves, Meta would not simply lose a prominent researcher. It would also lose a leading public voice associated with a research direction that is not centered on marketing today’s LLMs as the answer to every problem.
What the reported exit signals
LeCun has been openly skeptical about how AI technology, especially LLMs, is being marketed as a cure for all of humankind’s ails. He has also said AI systems still have a long way to go.
That skepticism is important context for the reported startup plan. A company focused on world models would fit with a view that today’s dominant AI systems are not the final form of the technology.
For Meta, the timing would be difficult. The company is investing heavily, recruiting aggressively, and reorganizing around a new AI unit while still maintaining a long-term research division. A LeCun departure would make that internal balance more visible.
For the wider AI market, the report points to continued movement of senior researchers, capital, and ideas between big technology companies and startups. World models are already being pursued by Google DeepMind and World Labs. A LeCun-led startup would add another high-profile effort to that field, if the reported plans move forward.
The immediate takeaway is measured but significant: Meta may be facing the loss of a renowned AI scientist at the same time it is trying to prove that its AI reset can close the gap with OpenAI, Google, and Anthropic.