Mark Zuckerberg’s Meta is expanding its Superintelligence Lab with three hires from AI security company

Mark Zuckerberg’s Meta is strengthening its defenses as the global artificial intelligence (AI) race hots up. The social media giant’s billion-dollar AI division, Meta Superintelligence Labs, has reportedly hired the three co-founders and key team members of prominent AI security startup Virtue AI.

According to a report by Axios, Meta is bringing in the founders as well as a broader team from the startup.The talent acquisition comes at a time of global anxiety surrounding AI safety. The industry is currently reeling from national security crackdowns that forced Anthropic to pull back its cutting-edge models over fears that their built-in guardrails could be disabled.As tech giants transition from basic chatbots to autonomous “AI agents” capable of executing complex real-world actions, keeping these tools secure has become an absolute necessity.

Meta sourcing elite security talent

According to an internal Meta memorandum viewed by Axios, the high-profile hires include Virtue AI co-founders Bo Li, Dawn Song and Sanmi Koyejo. All are recognised as premier leaders in the field of AI security. Virtue AI specialises in developing critical enterprise-grade security tools. These include automated red teaming which actively hack their own AI models to find hidden vulnerabilities before malicious actors do.

They also develop real-time safety nets that block an AI from generating harmful or illegal output mid-conversation, and create structural frameworks to ensure algorithms comply with strict legal and ethical standards.The move underscores an internal memo circulated within Meta regarding the high stakes of automated software. “As we ship AI products to billions of people and build increasingly capable agents, keeping those systems safe, reliable, and trustworthy is foundational,” the company memo reportedly stated.Li and Song will report directly to Nat Friedman, a top executive within the Meta Superintelligence Lab. Meanwhile, Koyejo will report to Rob Fergus, the head of Meta’s Fundamental AI Research (FAIR) laboratory.Meta deployed this acqui-hire strategy just months ago in March, when it successfully poached the founders and staff behind Dreamer, an AI startup dedicated to agent construction.

Tech giants push for ‘Agentic’ security

Global AI laboratories are rapidly doubling down on “agentic” security due to intense government scrutiny. Silicon Valley is currently navigating a strict regulatory environment where the White House is willing to utilise emergency export controls to pull advanced software offline over national security reasons. While Anthropic struggles to convince regulators that its Fable 5 and Mythos models are safe to return to the public sphere, other rivals are also feeling the heat.