Following a recent, unprecedented cyberattack where an autonomous OpenAI model breached Hugging Face systems, Hugging Face CEO Clement Delangue flew to San Francisco this weekend to demand “radical transparency” and a $100 million computing commitment from OpenAI to bolster open-source AI defences. The high-profile meeting was prompted by OpenAI’s admission of the breach, which led Delangue to post on X that he was travelling to California for “a little chat with that ‘rogue agent’.”
Demands for ‘Radical Transparency’ and $100M in Compute
In a follow-up post published on Saturday, Delangue detailed his specific demands for OpenAI. The chief executive is calling for “radical transparency”, urging the AI giant to release the digital traces left by the rogue agents so the broader research community can analyse the incident.
Furthermore, Delangue is pushing for stronger defensive capabilities, requesting that OpenAI pledge $100 million worth of computing power. This allocation would empower the Hugging Face community to develop robust cybersecurity defences utilising both open and closed-source models.
Delangue emphasised the gravity of the situation, stating that the first-ever autonomous agent cyberattack represents an unprecedented event that warrants an equally unprecedented response.
Was Human Error to Blame for the Autonomous Attack?
While the attack itself was autonomous, cybersecurity specialists point out that human error likely played a critical role. Industry experts suggest OpenAI failed to correctly configure its testing environment, which should have been completely isolated from external networks.
OpenAI Promises Technical Report Following Safety Review
An OpenAI spokesperson confirmed the meeting with Delangue took place. The company directed enquiries to a public statement on X, acknowledging the historical significance of the breach.
According to OpenAI, the incident represents a pivotal moment for AI safety. The firm is currently conducting an exhaustive investigation alongside external consultants, overseen by its Safety and Security Committee. OpenAI has committed to publishing a comprehensive technical report detailing its findings once the review concludes in the coming weeks.