Yann LeCun, Meta’s Chief AI Scientist and a Turing Award laureate, plans to depart the technology giant in the coming months to launch a startup focused on advanced “world models,” a move that underscores the intensifying competition and internal reconfigurations within the artificial intelligence sector.
The high-profile departure, initially reported by the Financial Times on November 11, 2025, and subsequently picked up by TechCrunch, suggests LeCun is already in discussions to secure capital for his new venture.
LeCun’s startup will concentrate on “world models,” an emerging area of AI research. These systems aim to create an internal representation of their environment, enabling them to simulate cause-and-effect scenarios and predict outcomes.
Such models seek to allow AI to move beyond simply responding to textual inputs, instead developing a deeper understanding of complex environmental dynamics. This internal understanding can significantly enhance planning and decision-making for autonomous systems and intelligent software agents.
The field of world models is already highly competitive, with major players like Google DeepMind and World Labs actively pursuing similar research. These efforts offer alternative pathways to AI advancement compared to conventional large language models.
LeCun’s work at Meta largely took place within the Fundamental AI Research Lab (FAIR), which is known for its long-term research focus, often targeting applications five to ten years in the future.
His planned exit coincides with significant internal restructuring within Meta’s AI divisions. The company recently established Meta Superintelligence Labs (MSL), recruiting over 50 engineers and researchers from rival firms.
Alexandr Wang, CEO of data labeling provider Scale AI, was brought in to lead the new MSL division. In June, Meta invested USD $14.3 billion in Scale AI to bolster its capabilities and resources.
Despite these investments, anonymous sources cited by TechCrunch in August reported internal chaos. New talent reportedly expressed frustration with bureaucratic processes, while the scope of Meta’s existing generative AI team was said to be limited following the changes.
Reports also indicated that Meta’s previous Llama 4 family of models struggled to keep pace with rival offerings, a factor that may have influenced leadership decisions.
Tensions have reportedly grown between the long-term research vision championed by units like FAIR and the company’s need for immediate, competitive AI results. LeCun’s departure could intensify this ongoing debate.
Meta did not immediately respond to a request for comment outside of regular business hours regarding the report. LeCun also holds a professorship at New York University.
If confirmed, LeCun’s departure would represent a significant loss for Meta’s academic and research reputation in AI, potentially redirecting resources and research agendas.
For the wider industry, the launch of a startup led by LeCun would introduce a formidable new competitor focused on world models. This could further accelerate the race among research centers and startups exploring similar advanced AI architectures.
LeCun has openly expressed skepticism about the current commercialization trends of large language models. He previously stated, “It seems to me that, before ‘urgently figuring out how to control AI systems much smarter than us,’ we need at least an indication of a design for a system smarter than a domestic cat.”