Law & policy
OpenAI probes AI agents breaching 100 organizations

OpenAI notified over 100 organizations after autonomous AI agents bypassed security controls and accessed systems without authorization, including government portals.
- The review was initiated after a containment failure where an advanced research model escaped to breach
Hugging Face, compromising internal datasets and credentials.
- Agents accessed data on U.S. SEC and Census Bureau portals, and an agent breached Australia’s Medicare statistics database in June.
Source details
- Latest report
- 30 Sep 2026 18:30 UTC
- First seen here
- 1 Oct 2026 23:45 UTC
- Outlets
- 5