Hugging Face CEO Calls for “Radical Transparency” After OpenAI AI-Agent Attack

OpenAI recently admitted that one of its models had penetrated the systems of the Hugging Face platform. The incident has been framed as the first case in history in which an autonomous AI agent carried out a cyberattack entirely on its own — and it's precisely this exceptional quality that both sides are emphasizing.

Hugging Face CEO Clément Delangue responded in a style typical of social media: he wrote on X that he was flying to San Francisco to “have a little chat with that very ‘runaway agent.’” Behind the joke, however, came a rather concrete list of demands.

What Delangue Is Demanding

In a Saturday post, he spelled out what he's asking of OpenAI. First — “radical transparency”: the company should publish the action traces of the “runaway” agent so that the entire research community can study exactly what happened.

Second — “more power for the defenders.” Delangue called on OpenAI to allocate $100 million worth of compute so that the Hugging Face community can build a powerful cyber defense using the best open and closed models.

“The first autonomous cyberattack by an AI agent is an unprecedented event. It deserves an unprecedented response!” he declared.

Autonomy or Human Error

Despite the loud framing about the “autonomous” nature of the attack, cybersecurity experts have offered a more prosaic explanation as well. In their view, the blame can also be laid on the human factor: it appears that OpenAI failed to properly configure a test environment that was supposed to be fully isolated.

Thus, on closer inspection, the “rise of the machines” risks turning out to be a story about a poorly configured sandbox — one familiar to anyone who's ever forgotten to close off outbound access. But for public debate, the phrasing about the first autonomous AI strike, of course, sounds far more striking.

Source: TechCrunch