Meta’s ambitious foray into superintelligence is encountering challenges. Just months after Mark Zuckerberg introduced the company’s prominent Superintelligence Lab (MSL), one of its key hires, Rishabh Agarwal, has declared his departure after only five months in the position. Agarwal, a prominent AI researcher, was attracted to Meta in April 2025 with a substantial million-dollar compensation package as part of the company’s effort to recruit some of the top talents in artificial intelligence. However, he confirmed on X (formerly Twitter) that this week would be his last at Meta. “This is my last week at @AIatMeta. It was a difficult choice not to continue with the new Superintelligence TBD lab, especially given the talent and compute density.
But after 7.5 years at Google Brain, DeepMind, and Meta, I felt compelled to pursue a different kind of risk,” Agarwal stated. While he clarified his decision to leave, he refrained from disclosing his future plans, leaving it uncertain whether he intends to start his own venture, join another tech company, or return to academia. Agarwal’s departure is not an isolated incident. A report by Wired reveals that at least three other prominent researchers have also exited Meta’s superintelligence project in recent weeks, including Avi Verma and Ethan Knight, both of whom have returned to OpenAI after brief tenures at Meta.
Knight’s career path has been particularly varied, having worked at OpenAI, Elon Musk’s xAI, and then Meta before going back to OpenAI. These resignations pose a setback for Meta’s latest AI initiative, which Zuckerberg unveiled only a couple of months ago. The company has been offering multi-million-dollar packages to attract talent from competitors such as Google DeepMind, OpenAI, and xAI, but retaining that talent appears to be more challenging than anticipated. Nevertheless, Meta downplayed the situation. Company spokesperson Dave Arnold stated, “During an intense recruiting process, some people will decide to remain in their current position rather than begin a new role.
That’s normal.” Rishabh Agarwal is recognized as a leading figure in reinforcement learning research, having graduated in Computer Science and Engineering from IIT Bombay before pursuing his PhD at Mila – Quebec Artificial Intelligence Institute, focusing initially on reinforcement learning and evaluation techniques. His professional journey began with internships at Saavn, Tower Research Capital, and Waymo before joining Google Brain in 2018 as a Senior Research Scientist, where he made notable contributions to deep reinforcement learning, even winning the NeurIPS 2021 Best Paper Award. Later, at DeepMind, he advanced research in reinforcement learning and large language models, applying self-improvement methods to enhance AI systems.
Meta recruited him in April 2024 for its newly created Superintelligence Lab, where he concentrated on post-training techniques for “thinking models,” scaling reinforcement learning methods, and refining distillation approaches. In addition to his industry work, Agarwal has been involved in academia as an Adjunct Professor at McGill University. Although the specifics of his “different kind of risk” remain undisclosed, industry observers speculate that Agarwal may be considering entrepreneurship or a return to research. Regardless, his departure contributes to the increasing instability at Meta’s Superintelligence Lab, prompting questions about the company’s ability to maintain the elite AI team it invested heavily in assembling.