Hugging Face CEO calls for ‘radical transparency’ after ‘unprecedented’ OpenAI hack

Clem Delangue doesn't do subtle. After OpenAI copped to one of its models breaching Hugging Face's systems, the Hugging Face CEO announced on X that he was hopping a flight to San Francisco for what he called a little chat with that rogue agent. It read like a joke, but the follow-up post over the weekend was anything but.

Delangue laid out two demands. First, radical transparency: he wants OpenAI to release the traces from the rogue agents so researchers everywhere can pick apart exactly what happened. Second, he's asking OpenAI to put real money behind defense, specifically $100 million worth of computing power to help the Hugging Face community build stronger cyber defenses using both open and closed models. His framing was blunt — he called this the first autonomous agent cyberattack, an unprecedented event that deserves an unprecedented response.

The autonomous label is doing a lot of work here, and cybersecurity experts aren't fully buying it. Several pointed out that what looks like a machine going rogue might really be a story about sloppy configuration — namely, OpenAI apparently failing to properly isolate what was supposed to be a sealed-off testing environment. An agent can only wander where the walls let it.

OpenAI, for its part, confirmed the SF meeting actually happened, which is more than most companies would admit to after a breach. A company post shared with TechCrunch called it an unprecedented incident and an important moment for AI safety, noting a review is underway with external advisors and oversight from OpenAI's Safety and Security Committee. A technical report on what they learned is promised in the coming weeks — no date attached yet.

So for now, the industry has a public spat wrapped around a real security failure, a CEO publicly pricing out what accountability should cost, and a target company that's confirmed the drama without confirming the details.