Meta Hires Three Virtue AI Founders Into Superintelligence Labs

Meta Superintelligence Labs, the company's frontier AI unit, now has the team behind one of the more credible independent AI security startups working inside it. According to Axios, Meta has absorbed three co-founders of Virtue AI, Bo Li, Dawn Song, and Sanmi Koyejo, along with additional members of the startup's team. Virtue AI built enterprise AI security infrastructure covering automated red teaming, runtime guardrails, and AI governance tooling.

The reporting lines say something about how Meta is integrating this work. Li and Song will report to Nat Friedman at Meta Superintelligence Labs, placing them alongside frontier model development. Koyejo will report to Rob Fergus, head of FAIR, Meta's fundamental AI research lab, giving both the product-focused and research-focused sides of Meta a direct link to adversarial evaluation expertise.

An internal memo framed the rationale plainly: "As we ship AI products to billions of people and build increasingly capable agents, keeping those systems safe, reliable, and trustworthy is foundational." The financial terms of the arrangement were not disclosed.

What the reporting does not give you is clarity on what happens to Virtue AI's existing enterprise customers, or how the reporting split between two distinct Meta units will work in practice when coordination is needed. Concentrating red-teaming expertise inside the company whose models they are evaluating is a structural tension worth watching regardless of how it is managed.

What Meta gets, if this holds together, is adversarial evaluation capability built by researchers who developed it independently. That independence of perspective is presumably what made Virtue AI worth bringing in. Whether it survives the move inside a company as large as Meta is the open question this deal raises.