Meta reportedly cuts the Virtue AI team it hired in June

Meta has reportedly fired the Virtue AI employees it hired in June, according to a post circulating on X. If accurate, the move would undo one of the company’s more visible AI safety hires only months after it happened.

The claim has not been independently confirmed. Research into the matter turned up no substantial evidence of a layoff, and coverage of the team has so far centered on the hiring itself.

Who Meta brought in, and why

On June 25, 2026, Meta hired key personnel from Virtue AI, including all three of its co-founders: Bo Li, Dawn Song, and Sanmi Koyejo. The team was folded into Meta Superintelligence Labs, the unit housing the company’s most ambitious AI work.

The arrangement was described as an acqui-hire. Meta was buying the talent, not the company.

The expertise Meta was after fell into three buckets: automated red teaming, runtime guardrails, and AI governance.

Red teaming means attacking your own system on purpose to find weaknesses before someone else does. Automating it lets software run those attacks at scale, probing a model far faster than a room of human testers could.

Runtime guardrails are the bouncers. They sit between a live AI model and its users, catching harmful or off-policy outputs as they happen rather than after the fact.

AI governance covers the policies, audits, and controls that determine how a company builds and deploys models, especially as regulators pay closer attention.

Virtue AI itself was young. Founded in 2024, the startup raised $30 million in funding and counted Uber and Glean among its clients.

A busy year of cuts and deals

The reported firings would land against a backdrop of heavy restructuring at Meta. In May 2026, the company laid off around 8,000 employees, approximately 10% of its workforce.

Those cuts were tied to a reorganization of AI projects under an initiative called Project OT. They came before the Virtue AI hires and are not reported to be connected to them.

Meanwhile, what was left of Virtue AI found another buyer. Fortinet announced its acquisition of Virtue AI on August 17, 2026, with the stated aim of strengthening its AI-native security offerings.

What this means for AI safety and Big Tech hiring

If the X post proves accurate, the most immediate question is what it says about Meta’s commitment to AI safety work inside Superintelligence Labs. The Virtue AI hires were framed as part of an effort to harden the safety measures around Meta’s AI deployments.

It is also possible the post is wrong, partially right, or describing something more nuanced than a mass firing. Until Meta or the individuals involved confirm the details, the story remains an allegation rather than an established fact.

Acqui-hires have become a common way for large companies to absorb specialized teams without taking on an entire startup. Without the product, customers, or company attached, the talent is the whole deal.

Fortinet’s purchase points to a different appetite. Cybersecurity firms appear eager to bolt AI safety capabilities onto their existing products, treating tools like red teaming and guardrails as natural extensions of traditional security.

Bo Li, Dawn Song, and Sanmi Koyejo were brought in to help Meta keep its AI in check. Whether they are still there to do it is, for now, an open question.