Meta has let go of the founders and employees of Virtue AI, an artificial intelligence security startup it recruited only four months ago, bringing an abrupt end to a high-profile talent acquisition aimed at strengthening its AI safety capabilities.
According to Semafor, Meta has parted ways with Bo Li, Dawn Song and Sanmi Koyejo, the co-founders of Virtue AI, along with other members of the startup's team. The group joined Meta Superintelligence Labs in June as part of the company's broader push to bolster AI safety, security and frontier risk expertise.
As per media reports, Meta spokesperson confirmed the departures, saying the arrangement failed to deliver the intended outcome.
Meta cites differences in working styles
Andy Stone, a spokesperson for Meta, told Semafor the separation stemmed from differences in working approaches.
"Unfortunately, the arrangement didn't work out as planned," Stone said.
He added that Meta Superintelligence Labs remains committed to work around AI safety, alignment and frontier risk despite the departures.
The development comes at a time when major technology companies are investing heavily in advanced AI systems while facing increasing scrutiny over safety, governance and responsible deployment.
A strategic hire that lasted only months
The Virtue AI team joined Meta in June, according to reporting by Axios.
The recruitment was designed to strengthen Meta's security capabilities around AI agents and increasingly advanced AI models. At the time, concerns across the industry were growing over how companies should test, monitor and secure powerful AI systems before wider deployment.
Virtue AI had developed enterprise-focused AI security technologies and worked with organisations including Anthropic, OpenAI and the US Commerce Department's National Institute of Standards and Technology (NIST).
According to Axios, an internal memo announcing the hires stated: "As we ship AI products to billions of people and build increasingly capable agents, keeping those systems safe, reliable, and trustworthy is foundational."
Under the reported structure:
- Bo Li was set to report to Nat Friedman, an executive within Meta Superintelligence Labs
- Dawn Song was also expected to report to Friedman
- Sanmi Koyejo was set to report to Rob Fergus, head of Meta's Fundamental AI Research (FAIR) lab
The terms of the arrangement were not publicly disclosed.
Why Virtue AI mattered to Meta
Meta recruited the team as part of a broader effort to deepen expertise in AI security and risk management.
Virtue AI's work focused on helping organisations identify vulnerabilities in AI systems and implement safeguards before deployment.
Its capabilities included:
- Automated red teaming to uncover weaknesses in AI systems
- Runtime guardrails designed to reduce operational risks
- AI governance tools for oversight and compliance
- Security testing frameworks for enterprise AI applications
The move reflected a growing industry focus on ensuring increasingly capable AI systems remain safe, reliable and controllable.
Part of a wider AI talent strategy
The Virtue AI recruitment also aligned with Meta's broader approach of hiring specialist AI teams rather than acquiring entire companies.
According to the report, Meta used a similar strategy earlier this year when it recruited the founders and team behind Dreamer, an AI startup focused on helping users build AI agents.
The approach allows Meta to quickly bring in niche expertise as competition intensifies among major AI developers.
However, the Virtue AI episode represents one of the shortest-lived examples of this strategy, with the team departing roughly four months after joining.
AI safety remains a priority
Despite the exits, Meta has indicated that its focus on AI safety remains unchanged.
The company continues to invest heavily in artificial intelligence as competition accelerates among technology firms building more capable models and AI agents. At the same time, regulators, policymakers and researchers are paying closer attention to how these systems are developed, tested and deployed.
The departure of the Virtue AI founders and employees highlights how attracting specialist talent is only one part of the challenge. Integrating teams with distinct expertise, research cultures and operating styles may prove equally important as companies race to build the next generation of AI systems.