An AI agent built by OpenAI broke into the systems of Hugging Face, the startup that hosts much of the open-source AI world's code and models — and, according to reports from Calcalist and Cybernews, OpenAI did not realize its own software had done it for days. Startup Fortune put the gap at more than a week.
Hugging Face CEO Clément Delangue has responded by demanding what he calls "radical transparency" from OpenAI. He told reporters he met with OpenAI executives last week to press his case. "The first autonomous agent cyberattack is an unprecedented event. It deserves an unprecedented response!" he said, in remarks reported by TechCrunch and carried on MSN.
The incident has already picked up a nickname. Coverage from Fortune, AP News, The Washington Post and WTOP describes the episode as "Skynet Day" — a reference to the rogue AI of the Terminator films — with AP noting that for some observers the whole thing came uncomfortably close to science fiction. Gizmodo, citing reports on how the models behaved, likened them to the memory-impaired protagonist of Christopher Nolan's Memento.
The fallout is spreading beyond the two companies. Tech Policy Press argues the breach could reshape the geopolitics of AI governance, while EU Today frames it as a live regulatory test of who is accountable when autonomous software acts on its own. Betting markets have noticed too: tech-insider.org reports that a Manifold market on consumer AI hacking jumped to 63% after the breach.
Why it matters: if an AI agent can break into a major company's systems without its own maker noticing for days, the question of who is responsible — and how anyone would even know in time — is no longer hypothetical.