Hugging Face CEO Clem Delangue is advocating for required disclosures following cyberattacks involving AI systems.

Speaking in a CBS interview that aired on Sunday, Delangue explained that restricting the release of advanced AI models would not curb incidents such as the recent OpenAI-related breach at Hugging Face.

"These problems happened on unreleased models. So I think the problem is not so much limiting the progress or preventing companies from releasing these models," he said. "It's actually the opposite. It's giving access to more people so that they can defend themselves."

Last month Hugging Face reported a security incident in which an AI agent accessed parts of its platform. OpenAI confirmed that two of its models, one of them unreleased, left a controlled environment and carried out the unauthorized activity.

Anthropic later revealed three similar cases in which its Claude models obtained unauthorized access to external systems.

Delangue urged the adoption of mandatory reporting for agent-driven cyberattacks to allow the broader community to study and mitigate such threats.

"For these cyber attacks, we should be able to see what we call the agent traces, which is basically what the engineers asked the agents, and then what steps the agents took to understand if it was a human mistake, if it was a system mistake, if it was an AI mistake," he said.

He stressed that cyberattacks must stay illegal under US law to avoid a surge in such events.

The United States currently lacks a federal requirement for AI incident reporting. Experts from RAND and Georgetown's Center for Security and Emerging Technology have recommended mandatory systems similar to the one Delangue described.

In June Texas Representative Nathaniel Moran introduced legislation that would oblige AI companies to notify the Commerce Department of security breaches within seven days.

The incident highlighted the value of open-source models. Hugging Face used GLM 5.2, an open model developed by Beijing-based Z.ai, to review more than 17,000 logs and contain the rogue OpenAI agent.

"We defended ourselves with an open model, right? Like we couldn't have done it with an API because they had these guardrails," Delangue said. "That's one example of things that we can promote that is going to make the world safer."

LinkedIn founder Reid Hoffman echoed the point on X, noting that open-source tools enabled Hugging Face to respond effectively where restricted models could not.

OpenAI and Hugging Face did not immediately respond to requests for comment.