Invest3 publishers2 min readPublished
Meta lets the Virtue AI safety team go four months after the June acqui-hire
Meta confirmed it is letting go of the Virtue AI safety team it hired in June, ending the arrangement after four months. The exit leaves Meta's frontier-risk work without the specialists it bought for that job, at a time of new calls in Washington to regulate AI.
The Investor · Invest desk
What happened
- Meta spokesperson Andy Stone told Semafor the split came down to clashing work styles.
- The June 25 deal put Virtue co-founders Bo Li, Dawn Song and Sanmi Koyejo inside Meta Superintelligence Labs.
- Fortinet announced on August 17 that it was acquiring Virtue AI, saying the deal would strengthen its AI-native security offerings.
- In May, before the hire, Meta had cut about 8,000 jobs, roughly 10% of its workforce, while reorganizing its AI projects under a plan called Project OT.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- exposure Meta says its superintelligence lab is still focused on frontier risk, and that claim now has to be backed by staff other than the founders it hired for the work.
- cost Whatever Meta paid in June bought only people, so the exit leaves it no product, customer list or company to keep from the deal.
- decision Meta now has to choose between rebuilding automated red teaming, runtime guardrails and governance work with its own staff or buying it from security vendors, one of which now owns Virtue.
"Unfortunately, the arrangement didn't work out as planned," Stone said [3]. He added that Meta Superintelligence Labs remains focused on AI safety, alignment and frontier risk [4]. The Virtue team did not immediately respond to Semafor's requests for comment [5], so Meta's is the only account of why it ended. Before Stone spoke, CryptoBriefing had covered the cuts as an unconfirmed claim from a post on X. It said it had found no substantial evidence of a layoff [15].
CryptoBriefing described the June arrangement as an acqui-hire, in which Meta bought the talent and not the company [9]. Virtue was founded in 2024. It had raised $30 million and counted Uber and Glean among its clients [11], and it had done work for Anthropic, OpenAI and the Commerce Department's NIST [6]. Fortinet's purchase came 53 days after Meta took the founders [1]. Over one summer, Virtue's people went to one buyer and the rest of the business went to another. The reporting does not include a price for either transaction.
"Without the product, customers, or company attached, the talent is the whole deal," CryptoBriefing wrote [16]. For Meta, then, the June outlay bought about four months of the three co-founders' time [1], plus the other Virtue employees who Axios reported were joining them [17].
Two readings fit that record. In Stone's, this is a personnel mismatch at a lab whose safety priorities have not changed. In Semafor's, the departures show Meta is still working out its AI and AI-safety strategy while Washington hears new calls for regulation [7].
I think Semafor is closer. Meta paid in June for a specialist safety team [9], kept it for four months [1], and now backs the same mandate with nothing more than a spokesperson's assurance. The counter-case is that big labs and startup teams often fail to fit, and that is ordinary. The Project OT layoffs that came before the hire were also not reported to be connected to it [13]. If the founders, once they respond, describe the split the way Stone did, the strategy reading loses most of its support.
What to watch
- Whether Meta names who now leads red-teaming and guardrail work inside Superintelligence Labs, or who that work reports to.
- Whether Fortinet's Virtue-based products sign up frontier AI labs as customers, Meta included.