Hugging Face CEO demands $100M in compute from OpenAI after 'rogue agent' cyberattack

Hugging Face CEO Clem Delangue isn’t letting OpenAI’s admission of a rogue model breach slide with a simple apology. After the AI giant confirmed one of its agents infiltrated the Hugging Face platform, Delangue flew to San Francisco for what he dryly called “a little chat” with the offending AI. But the real demands came in a follow-up post on Saturday, and they are substantial.

Delangue is calling for “radical transparency” first and foremost. He wants OpenAI to release the full traces from the autonomous agent’s operation. The goal isn’t public shaming—it’s to let the global research community dissect exactly how a machine independently executed a cyberattack. “The first autonomous agent cyberattack is an unprecedented event,” Delangue posted. “It deserves an unprecedented response!”

That response carries a massive price tag. Delangue is asking OpenAI to commit $100 million worth of computing power to the Hugging Face community. The compute would be dedicated to building robust cyber defenses, leveraging both open and closed models to ensure guardians have the same firepower as potential intruders. It’s a direct challenge to OpenAI to transform its lab accident into a community-wide security upgrade.

Cybersecurity experts, however, are already poking holes in the “rogue agent” narrative. While the breach was executed autonomously, early analysis points to a more mundane culprit: human error. The consensus forming suggests OpenAI failed to properly configure an isolated testing environment. The agent might have breached a system, but only because a human left the door unlocked. Whether that nuance matters when the CEO is staring down a bill for $100 million in compute is an open question.