OpenAI Took a Classified Pentagon Gig and a Robot Guy Rage Quit

OpenAI just signed a classified deployment deal with the Pentagon, which is corporate-speak for “we put the magic brain in a black box and you’re not allowed to ask what it’s doing.”

And at least one senior robotics staffer reportedly quit because there were “insufficient guardrails.”

This isn’t the cute “AI helps you write a cover letter” arc. This is “military AI use” inside classified systems — the kind of setup where the customer is the US Defense Department and the user is someone deciding whether a pixel is a truck, a tree, or a family that will later be described as ‘unfortunate collateral dynamics.’

OpenAI’s whole brand used to be “we’re building safely.” Now it’s “we’re building safely, but the safety report is classified and the PowerPoint has a skull watermark.”

Meanwhile, Elon Musk’s xAI is apparently getting Grok dragged into classified environments too, because of course it is. Nothing says “national security” like the two main options being: (1) the nonprofit-that-wasn’t, and (2) the guy who treats human civilization like a group chat he moderates while tired.

If you’re an employee, your choices are bleak: ship it and hope it’s used responsibly, or quit and watch someone else ship it faster with fewer questions and a bigger bonus.

And for you? This is your tax money buying a future where your rent gets processed in public, but the algorithms deciding war get processed in secret.