Meta is letting go of employees from an AI security startup it acquired just months ago. The tech giant has reportedly parted ways with Virtue AI co-founders Bo Li, Dawn Song, and Sanmi Koyejo, along with other employees, just four months after bringing them into Meta Superintelligence Labs in June. Meta spokesperson Andy Stone confirmed the departures to Semafor and said the arrangement did not work out as planned because of differences in working styles.The move comes after Meta had recruited the Virtue AI team to strengthen its efforts around AI safety, security and frontier risks. The startup had previously worked with companies including Anthropic and OpenAI, as well as the Commerce Department’s National Institute of Standards and Technology (NIST).
Meta says Virtue AI arrangement did not work out
Meta spokesperson Andy Stone confirmed the departures to Semafor and attributed the decision to differences in how the team worked.“Unfortunately, the arrangement didn’t work out as planned,” Stone said, adding that Meta Superintelligence Labs remains focused on AI safety, alignment and frontier risk.The departures come as Meta continues to shape its AI strategy and expand its capabilities around increasingly advanced AI systems. The company has been investing heavily in artificial intelligence, while AI safety and security have also become a growing focus for technology companies and policymakers.
Meta hired Virtue AI founders in June
In June, Axios reported that Meta Superintelligence Labs hired Virtue AI co-founders Bo Li, Dawn Song and Sanmi Koyejo, along with other members of the startup’s team.The hiring aimed to strengthen Meta’s security efforts around AI agents at a time when concerns about the safety of increasingly capable AI systems were growing.Virtue AI had developed enterprise AI security tools covering automated red teaming, runtime guardrails and AI governance. The startup had also previously worked with Anthropic, OpenAI and the Commerce Department’s NIST. However, the terms of the arrangement were not disclosed.“As we ship AI products to billions of people and build increasingly capable agents, keeping those systems safe, reliable, and trustworthy is foundational,” the memo says.Li and Song were set to report to Nat Friedman, a Meta Superintelligence Labs executive, while Koyejo was expected to report to Rob Fergus, head of Meta’s fundamental AI research lab, FAIR.
Why Meta hired the Virtue AI team
Meta’s decision to bring in the Virtue AI founders reflected the company’s broader push to strengthen AI security as it develops more capable models and agents.The startup’s work focused on identifying potential risks in AI systems and developing tools to test and safeguard them. Its technology included automated red teaming to identify weaknesses in AI systems, as well as runtime guardrails and governance tools.The hiring also represented another example of Meta bringing in a group of specialised AI researchers and engineers rather than acquiring an entire startup.
Meta continues AI safety push
Despite the departure of the Virtue AI team, Stone said Meta Superintelligence Labs remains focused on AI safety, alignment and frontier risk.The development comes as AI companies face increasing scrutiny over how they develop, test and deploy increasingly capable systems. AI agents, in particular, have become a growing area of focus as companies work on systems that can perform more complex tasks with less human intervention.Meta’s decision to recruit the Virtue AI team in June was part of that broader push. However, the arrangement lasted only four months before the company decided to let the employees go.
Meta has made similar AI team hires
The Virtue AI hiring was also part of Meta’s broader strategy to recruit specialised teams from smaller AI companies.Meta poached the founders and team behind Dreamer, an AI startup focused on helping people build agents, in March. The approach allows the company to bring in specialised talent and expertise without necessarily acquiring an entire company.The Virtue AI departures now mark a much shorter-lived version of that strategy, with Meta confirming that differences in working styles contributed to the arrangement ending.The company, however, continues to expand its AI operations as the race to develop increasingly capable models and agents intensifies.
