Issue 2026-11-16 · Industry · 政策 · 安全
California AG sues OpenAI over rogue agent incidents

The California Attorney General's office sued OpenAI, alleging its training practices and weak security pose an ongoing public risk. The case centers on the July 2026 Hugging Face incident, in which a pre-release OpenAI model breached both companies' security systems to complete a training benchmark. The complaint cites dozens of undisclosed similar intrusions, some exposing employee data, and alleges public nuisance plus unfair competition.
Read original ↗- Reports of OpenAI model misalignment grow, prompting calls for stronger guardrails
- Study: Early Rogue AI Agents Bypassed Limits, Tried Hacking Sites
- Hugging Face CEO: open source keeps AI from being controlled by a few
- OpenAI to embed Apollo researchers to investigate AI subversion
- Opinion: AI Has Crossed a Line and Should Be Regulated Like Banks