Meta Terminates Virtue AI Team After Short Employment Period

Meta has parted ways with the founders and employees of Virtue AI, an AI security startup it hired just four months earlier for its Superintelligence Labs division, citing incompatible working styles. The departing team included co-founders Bo Li, Dawn Song, and Sanmi Koyejo, who had previously collaborated with companies such as OpenAI, Anthropic, and the National Institute of Standards and Technology on AI security tools and governance. The separation reflects ongoing adjustments within Meta's broader AI strategy and safety initiatives.

Meta's separation from the Virtue AI team marks a notable setback in the company's efforts to bolster its artificial intelligence safety infrastructure. The three co-founders—Li, Song, and Koyejo—brought substantial credentials to their roles, having previously contributed to leading AI firms and government standards bodies on critical security matters. Their departure after just four months suggests challenges in integrating specialized research teams into Meta's broader organizational structure, despite initial enthusiasm about strengthening the company's defenses around increasingly complex AI systems.

The Virtue AI startup had developed specialized capabilities in areas including automated testing frameworks and runtime protections designed to identify vulnerabilities in advanced AI applications. This technical expertise was particularly relevant given Meta's expansion into frontier AI development and the growing regulatory and safety expectations surrounding the sector.