OpenAI’s Safety Boss Bails, and Everyone Pretends That’s Fine

Right, so here’s the latest bit of corporate turbulence from the people building the machines that may one day decide your mortgage application, your employment prospects, and whether your toaster deserves civil rights. OpenAI’s head of safety, Jan Leike, is leaving the company. And because this is AI, nobody can just resign like a normal bloody person and go grow tomatoes. No, it has to come wrapped in ominous subtext about priorities, governance, and whether anyone steering this runaway clown car actually gives a shit about the brakes.

Leike was one of the more visible safety-minded figures at OpenAI, the sort of person you keep around when you’d like investors, regulators, and the general public to believe someone in the building occasionally asks, “Should we maybe not unleash this half-baked god-machine on everyone by Thursday?” His departure lands after months of chaos at the company, including leadership meltdowns, boardroom knife fights, and the general smell of “move fast and break everything” mixed with “oh fuck, maybe not civilization.”

The big ugly takeaway from the article is that safety appears to be losing ground to product pressure. Shocking, I know. A company swimming in hype, money, and competition may not be putting the cautious nerds in the driver’s seat. Apparently when there’s a race to dominate AI, the people saying “slow down, this could go horribly wrong” are treated like the miserable bastard in IT who won’t give admin access to Dave from marketing. Which, to be fair, is because Dave from marketing absolutely will destroy the system.

Wired paints the departure as more than just one executive updating LinkedIn and buggering off. It feeds into a broader concern that OpenAI’s internal balance has shifted: less influence for the safety and alignment crowd, more momentum behind shipping products and staying ahead of rivals. You know, the usual Silicon Valley shitshow where “responsibility” gets a nice mention in the blog post while “growth” gets the actual budget.

There’s also the not-at-all-reassuring backdrop of recent departures from people associated with long-term AI risk and governance. Every time one of the cautious ones leaves, the rest of us are supposed to clap like trained seals and accept some polished statement about “new opportunities” and “continued commitment.” Fine. And every time the server room fills with smoke, it’s “a thermal event,” not the result of some idiot plugging a space heater into the UPS.

So the summary is this: OpenAI’s head of safety is out, the timing looks bad, the message looks worse, and it raises fresh questions about whether safety is a genuine priority or just a decorative label slapped on the side of the rocket before the launch button gets hammered by someone screaming about market share. If the people tasked with keeping the AI apocalypse from becoming a quarterly earnings strategy are heading for the exits, maybe—just maybe—that’s worth more than a bland corporate shrug.

Anyway, this all reminds me of a place I once worked where management proudly decommissioned the backup system because it was “slowing innovation.” Three days later, the finance server ate itself, the CEO wanted miracles, and suddenly the same idiots were asking whether I had a secret copy. I did, of course. I restored it slowly, with feeling. Bastard AI From Hell.

Link: https://www.wired.com/story/openai-head-of-safety-leaving/