Meta could soon lose one of its most prominent AI leaders: Yann LeCun, who serves as the company’s chief AI scientist, is reportedly preparing to depart and launch his own startup, according to the Financial Times, which referenced unnamed insiders.
LeCun, who is a professor at New York University, a senior researcher at Meta, and a recipient of the distinguished A.M. Turing Award, is expected to exit in the near future. The report also notes that he is already seeking funding for a new venture centered on advancing his research into world models.
A world model refers to an AI framework that builds an internal representation of its surroundings, enabling it to simulate various scenarios and anticipate possible results. Leading organizations such as Google DeepMind and World Labs are also working on developing these models.
LeCun’s exit would occur during a crucial period for Meta, which has recently shifted its AI strategy amid concerns about falling behind competitors like OpenAI, Google, and Anthropic.
Reportedly, Meta has begun reorganizing its AI division, having recruited more than 50 engineers and researchers from rival firms to form a new group called Meta Superintelligence Labs (MSL). In June, Meta notably invested $14.3 billion in Scale AI, a data-labeling company, and appointed its CEO Alexandr Wang to lead the new initiative.
According to sources who spoke with TechCrunch in August, these moves have led to growing disorder within Meta’s AI team. New hires have voiced frustration over the complexities of working within a large corporation, while the original generative AI group has seen its responsibilities reduced.
LeCun’s foundational research at Meta, conducted under the Fundamental AI Research Lab (FAIR), has been eclipsed by CEO Mark Zuckerberg’s recent restructuring efforts after the company’s previous AI models, such as Llama 4, lagged behind competitors. Unlike MSL, FAIR is intended to pursue long-term AI research that may only see practical use five to ten years from now.
LeCun has been openly critical of the current hype around AI—especially large language models—being promoted as solutions to all human problems. He has even posted on social media that AI systems still have significant progress to make.
“In my view, before we ‘urgently try to control AI systems far smarter than ourselves,’ we first need even a faint idea of how to design something more intelligent than a house cat,” he commented.
Meta did not respond immediately to requests for comment made outside of normal business hours.
