Hugging Face CEO says AI companies should be required to disclose hacks after OpenAI breach
Business Insider · 2026-08-03
Hugging Face CEO, Clément Delangue, advocates for mandatory disclosure of AI cyberattacks to enhance transparency and enable the AI community to collectively learn from and prevent future incidents, as reported on August 3, 2026. He argues that such transparency should extend to "agent traces," detailing engineer prompts and AI actions, to distinguish between human, system, or AI-driven errors. Delangue emphasizes that this approach, rather than restricting AI progress, allows broader access to models for defensive purposes. He cited Hugging Face's successful defense against a recent OpenAI breach using an open-source Chinese model, GLM 5.2, as proof of concept. Currently, there is no federal AI incident reporting law in the US, though researchers from RAND and Georgetown's Center for Security and Emerging Technology, along with Texas Rep. Nathaniel Moran, have proposed similar mandatory reporting systems. Delangue also stressed that cyberattacks must remain illegal under US law to prevent their proliferation.
*The full article also explores the implications of open-source models in cybersecurity and the ongoing debate surrounding AI regulation.*