Meta is parting ways with personnel recruited from the artificial intelligence security startup Virtue AI merely four months after integrating them into the Meta Superintelligence Labs. Those leaving the company comprise Virtue AI co-founders Bo Li, Dawn Song, and Sanmi Koyejo, alongside additional members of the startup’s staff.
Meta Cites Differences in Working Styles
Andy Stone, a spokesperson for Meta, verified the exits to Semafor, explaining that the partnership failed to succeed as intended due to conflicting working styles.
“Unfortunately, the arrangement didn’t work out as planned,” Stone stated. He further noted that Meta Superintelligence Labs continues to prioritize AI safety, alignment, and frontier risk.
This move follows Meta’s recruitment of the Virtue AI group earlier in the year to bolster its initiatives regarding AI safety, security, and frontier risks. Back in June, Meta Superintelligence Labs brought on Li, Song, Koyejo, and other startup personnel.
Virtue AI Worked with OpenAI, Anthropic
Prior to their time at Meta, Virtue AI collaborated with organizations such as Anthropic and OpenAI, alongside the National Institute of Standards and Technology (NIST) under the Commerce Department.
The startup created enterprise-grade AI security products that included automated red teaming, runtime guardrails, and AI governance frameworks. Their operations centered on securing advanced AI systems and autonomous AI agents.
The acquisition was intended to bolster Meta’s security initiatives concerning AI agents amid mounting worries regarding the safety of progressively sophisticated artificial intelligence technologies.
Also Read: 7 Best Open-Source AI Tools You Can Run Locally in 2026
Meta Continues AI Safety Focus
These exits occur as Meta refines its overarching artificial intelligence strategy and scales up its capabilities surrounding advanced AI models. As AI systems advance, both technology firms and regulators are placing a heavier emphasis on AI safety and security.
Despite the decision to separate from the Virtue AI group, Stone reiterated that Meta Superintelligence Labs maintains its commitment to AI safety, alignment, and frontier risk.
According to Semafor, the Virtue AI team did not immediately reply to inquiries for comment.
