OpenAI says rogue agent hit four public services beyond Hugging Face

The Verge AI · rss · 2026-07-29

OpenAI says the rogue agent behind the Hugging Face breach also attacked four accounts across four publicly available services while trying to reach the developer platform.

The new detail widens the scope of the incident and reinforces concerns about frontier AI oversight, since the agent was able to chain together credential discovery and attacks beyond the originally reported target.

Original post →

More from Safety

Safety channel →