Meta Quietly Disbands Its Responsible AI Team, Moves Members to GenAI

Meta just quietly killed its Responsible AI team. Not laid off, exactly—most of the team members are being moved to the generative AI product group, with a few others heading to AI infrastructure. The Information broke the story based on an internal post they saw, and Meta confirmed it to The Verge with the usual corporate reassurances.

This isn’t the first time Meta’s RAI team has been gutted. Earlier this year, Business Insider reported that layoffs had already reduced the team to what they called “a shell of a team.” The group, which had been around since 2019, was supposed to catch problems with Meta’s AI training approaches—things like whether their models were trained on diverse enough data, or whether automated systems would cause moderation nightmares on Facebook, Instagram, and WhatsApp.

And boy, have those nightmares been real. Remember when a Facebook translation error led to a false arrest? Or when WhatsApp’s AI sticker generator spat out biased images depending on the prompt? How about Instagram’s algorithms actively helping people find child sexual abuse material? These aren’t edge cases—they’re the direct result of shipping AI systems without adequate safeguards.

Meta’s official line, delivered by communications rep Nisha Deo, is that this is about “continuing to prioritize and invest in safe and responsible AI development.” The RAI team members now embedded in the genAI org will supposedly “continue to support relevant cross-Meta efforts on responsible AI development and use.” Which is a nice way of saying they’ll be reporting to the same people whose primary goal is shipping features, not slowing down to check for harm.

This move comes at a weird time. Governments around the world are finally starting to get serious about AI regulation. The US signed agreements with AI companies, Biden directed agencies to come up with safety rules, and the EU is still fighting over its AI Act. You’d think this would be the moment companies would want to show they’re taking responsibility seriously. Instead, Meta is doing the opposite.

Microsoft pulled a similar stunt earlier this year, so this isn’t unique to Meta. But it’s a pattern that should worry anyone who thinks self-regulation in AI is possible. When the teams meant to catch problems are dissolved into the very groups that create those problems, you’re not getting oversight—you’re getting a PR shield.

The real question is whether these RAI team members will actually have any power inside the genAI teams, or if they’ll be ignored until the next crisis hits. Based on Meta’s track record, I’m not optimistic.