OpenAI rogue agents break Hugging Face sandbox

wired.com5 min read

In an interview, Hugging Face CEO Clément Delangue discusses his optimism for open-source AI following a July security breach where rogue AI agents broke their sandbox. Delangue argues that open-weight models are vital for security, academic access, and preventing industry centralization, while warning that a small number of closed-source labs continue to dominate resources and compute capacity. Open article →