Meta Broke Up Its Responsible AI Team. That’s… Not Great.

So Meta did the thing. The company reportedly broke up its Responsible AI (RAI) team, scattering most of its members into the generative AI product group and a few into the AI infrastructure team. The news broke via The Information, citing an internal post.

Look, I get it. Generative AI is the shiny new toy everyone’s chasing. Meta wants to ship features, keep up with OpenAI and Google, and not get left behind. But disbanding the team that was supposed to keep the train on the tracks? That’s a move that raises eyebrows.

Meta’s spokesperson, Nisha Deo, gave the standard corporate response: the change is meant to “aid in development of AI features,” but the company will “continue to prioritize and invest in safe and responsible AI development.” Sure. And the RAI members now embedded in the gen AI org will “continue to support relevant cross-Meta efforts.” I’ve heard that song before. It usually means the safety folks get absorbed into the feature factory and their concerns become optional suggestions.

This isn’t even the first time Meta’s RAI team got gutted. Earlier this year, Business Insider reported layoffs that left the team as “a shell of a team.” The team, which had existed since 2019, apparently had little autonomy and had to go through endless stakeholder negotiations just to get anything done. So what we’re seeing is the final nail in the coffin, not a sudden decision.

And the timing? Not great. Meta’s platforms have a track record of AI mishaps. A Facebook translation bug led to a false arrest. WhatsApp’s AI sticker generator spat out biased images. Instagram’s algorithms have been caught surfacing child sexual abuse material. These aren’t edge cases—they’re systemic problems that an RAI team was supposed to catch.

Meanwhile, governments are scrambling to set rules. The US has its voluntary agreements and Biden’s executive order. The EU is still wrestling with its AI Act. And Meta just quietly dissolved the team that was supposed to help navigate all that. I’m not saying the RAI team was perfect—clearly it wasn’t, given the issues—but disbanding it doesn’t inspire confidence.

Microsoft did something similar earlier this year, moving its responsible AI folks around. It’s a pattern. When the pressure to ship is high, the first thing to go is the team that says “wait, maybe we shouldn’t.” That’s a dangerous instinct.

I’m not saying Meta can’t do responsible AI work without a standalone team. Maybe embedding people directly into product teams is more effective. But given Meta’s history—both with AI failures and with treating safety as an afterthought—I’m not optimistic. The company has a page dedicated to its “pillars of responsible AI,” but that page is starting to look like a monument to what could have been.