Yann LeCun, Chief AI Scientist at Meta and a prominent figure in artificial intelligence research, is reportedly preparing to depart the company to establish his own startup. The venture is expected to focus on advancing his work on "world models," according to a report by the Financial Times, which cited anonymous sources familiar with the matter.
LeCun, a professor at New York University and a recipient of the prestigious A.M. Turing Award, is anticipated to leave Meta in the coming months. Sources indicated he is already engaged in discussions to secure capital for the new enterprise. World models are defined as AI systems designed to develop an internal understanding of their environment, enabling them to simulate cause-and-effect scenarios and predict outcomes. Research into these models is also underway at other leading laboratories and startups, including Google DeepMind.
His reported departure comes as Meta navigates a significant overhaul of its AI strategy. The company has recently adjusted its approach to AI development amid competitive pressures from rivals such as OpenAI, Google, and Anthropic. Reports indicate Meta has restructured its AI organization, including the recruitment of over 50 engineers and researchers from competing firms, to build a new unit dubbed Meta Superintelligence Labs (MSL). In June, Meta also invested $14.3 billion in data-labeling vendor Scale AI and appointed its CEO, Alexandr Wang, to lead the new division.
However, internal accounts suggest these strategic shifts have introduced complexities within Meta’s AI unit. Sources told TechCrunch in August that new hires have expressed difficulties navigating the company's bureaucratic structures, while the scope of Meta’s existing Fundamental AI Research (FAIR) division has reportedly been curtailed. FAIR, where LeCun conducts his long-term research, typically focuses on AI techniques with a five-to-ten-year development horizon, a focus that has reportedly been overshadowed by CEO Mark Zuckerberg’s recent decisions following the performance of the company's Llama 4 AI models.
LeCun has publicly expressed skepticism regarding the current marketing of AI technologies, particularly large language models (LLMs). He has stated that present AI systems have considerable development ahead. In a social media post, he questioned the urgency of controlling highly intelligent AI systems, noting the absence of even preliminary designs for systems smarter than a "house cat." Meta did not immediately respond to requests for comment regarding the report.
