Meta hired a bunch of AI safety experts in June, fired them in September, and is now acting like this was just a normal “culture fit” issue. If you believe that, I’ve got a bridge to sell you.

The Timeline That Should Make You Uncomfortable

Here’s what actually happened. In late June, Meta brought on board the entire crew from Virtue AI – a legitimate safety startup that’s done work with Anthropic, OpenAI, and even NIST (that’s the National Institute of Standards and Technology, the government folks who set actual standards). We’re talking about Bo Li, Dawn Song, and Sanmi Koyejo, plus their team. These aren’t random hires off LinkedIn. These are people who know their stuff.

At the time, Meta sent around an internal memo – because of course they did – saying something about how “keeping systems stable and reliable is a fundamental requirement” as they roll out AI to billions of users. You know, the kind of corporate speak that sounds great until you realize it means absolutely nothing three months later when those same people are out the door.

And here’s the thing that really gets me: Meta specifically created the Meta Superintelligence Lab (MSL) for this work. They made a whole new lab. Then they… fired the people they hired to staff it? Or at least some of them?

The “Working Styles” Excuse

Andy Stone, Meta’s spokesperson, gave us this gem: “Regrettably, this arrangement did not work out as planned.” He’s blaming it on “a clash in working styles.”

Look, I’ve been covering tech for 15 years. “Working styles” is what you say when you don’t want to say the real reason. It’s the corporate equivalent of “it’s not you, it’s me” – except we all know it’s definitely you. It's the same instinct that led to a redaction fail exposing what Google's data center really uses, when the truth was supposed to stay hidden too.

“Regrettably, this arrangement did not work out as planned.”

What does that even mean? Did the safety experts want to actually, I don’t know, prioritize safety? Did they push back on shipping timelines? Did they ask uncomfortable questions about whether Meta’s AI agents were ready for billions of users?

Because if that’s the “clash,” then yeah, I can see how that wouldn’t work out. Not for Meta, anyway.

This Isn’t Really About Culture Fit

The timing here is what bothers me most. Three months. That’s barely enough time to figure out where the bathrooms are, let alone clash so badly on “working styles” that you need to let people go. Virtue AI wasn’t some random acqui-hire where Meta bought a startup for the tech and then quietly shuffled the founders into VP roles where they’d collect paychecks and do nothing. These people were brought in specifically for AI safety work.

And it’s not like Meta didn’t know who they were hiring. Virtue AI’s track record was public. Their approach to safety was known. If there was going to be a fundamental mismatch in how they operate, wouldn’t that have come up during, oh I don’t know, the hiring process?

Unless. And I’m just spitballing here. Unless the problem wasn’t that Meta didn’t know what they were getting. The problem was that they got exactly what they asked for, and then realized they didn’t actually want it.

The Pattern We Keep Seeing

This isn’t Meta’s first rodeo with this kind of thing. Big tech companies love to talk about responsible AI and safety and all that good stuff – right up until it slows them down or costs them money or makes them less competitive. Then suddenly it’s all about “agility” and “innovation” and “working styles.” It's the same pattern playing out at OpenAI, where its safety chief just quit amid complaints about a broken internal culture.

I mean, Stone made sure to add that MSL “remains dedicated to AI safety, alignment, and risk management.” Cool. Great. Super reassuring coming right after you just fired your AI safety team.

What This Actually Means

Here’s my take, and I don’t think I’m being cynical: Meta wanted the PR boost of hiring respected safety experts. They wanted to point to Virtue AI and say “see, we’re taking this seriously” while they race to ship AI products to billions of users. But when the reality of actually doing that work kicked in – when it meant maybe moving slower or being more careful or god forbid not shipping something – they decided the optics weren’t worth the hassle.

And now we’re left with Meta assuring us they’re still committed to safety, just without the safety experts they specifically hired for that purpose. Which is kind of like saying you’re still committed to your diet right after you fired your nutritionist and personal trainer because they kept telling you to eat vegetables.

The really frustrating part? Meta will probably be fine. They’ll keep shipping AI stuff. MSL will keep existing in some form. And in a few months, this whole thing will be forgotten except as a footnote in some longer article about AI safety theater in big tech. But it’s worth remembering what happened here – because it’s probably happening at other companies too, just more quietly.