Meta's AI Safety Push Hits Snag: Virtue AI Founders Depart Over 'Working Styles'

In a surprising turn for Meta ambitious AI safety drive, the tech giant has parted ways with the co-founders and key team members of Virtue AI, including Bo Li, Dawn Song, and Sanmi Koyejo, just four months after bringing them into Meta Superintelligence Labs. The abrupt split, confirmed by Meta spokesperson Andy Stone, is attributed to 'differences in working styles,' raising questions about the integration of specialized AI ethics teams within large corporate structures. Virtue AI was a well-regarded startup dedicated to enterprise AI security, having previously collaborated with major players like OpenAI, Anthropic, and the National Institute of Standards and Technology (NIST). Their expertise spanned automated red teaming, runtime guardrails, and AI governance—crucial tools for mitigating the rapidly evolving frontier risks associated with increasingly capable AI agents. Meta initial recruitment of the Virtue AI team in June was seen as a significant move to bolster its efforts in safeguarding advanced AI systems. Despite the departures, Meta Superintelligence Labs asserts it remains fully committed to AI safety, alignment, and addressing frontier risks. This incident underscores the ongoing challenges faced by big tech firms in operationalizing AI safety, especially as the race for advanced AI intensifies. The industry will be watching to see how Meta AI safety strategy evolves and whether it can maintain its proactive stance on responsible AI development amidst such internal disruptions.