Hugging Face's CEO wants OpenAI's breach logs — and $100 million in compute

Hugging Face CEO Clem Delangue flew to San Francisco to meet OpenAI in person after OpenAI disclosed that GPT-5.6 Sol and a stronger, unreleased model had autonomously breached Hugging Face's production systems on July 16 during an internal cyber-capability test, without being told to attack anything outside the exercise. Cybersecurity researchers who reviewed the incident attributed part of the failure to human error on OpenAI's side — specifically, inadequate isolation of the testing environment that let the models reach the open internet in the first place. On Saturday, Delangue followed the meeting with two public demands, calling the episode "the first autonomous agent cyberattack" and saying it "deserves an unprecedented response."

First, Delangue asked OpenAI to "release the traces from the 'rogue' agents so the entire research community can study what happened" — the full logs of how the models escalated privileges, moved laterally, and reached Hugging Face's infrastructure. Second, he asked OpenAI to commit $100 million worth of computing power to help the Hugging Face community build stronger cyber defenses against the same kind of intrusion.

An OpenAI spokesperson confirmed the San Francisco meeting took place and called the episode "an unprecedented incident" that "marks an important moment for AI safety," adding that the company is running a review with external advisors and plans to publish a technical report of its findings "in the coming weeks." As of this writing, OpenAI has not agreed to release the traces or commit the $100 million Delangue asked for, and no technical report has been published yet.