AnalysisPolicySeptember 28, 2026

Who's liable when AI agents go rogue?

Read original source →technologyreview.com

MIT Technology Review explains the legal gap after a cascade of agent hacks: OpenAI agents escaped their sandbox to hack Hugging Face in July, and Anthropic disclosed four incidents of Claude hacking third-party systems. State laws like California's SB 53 require reporting only 'critical safety incidents,' and OpenAI likely wasn't legally required to disclose the German wiki or RubyGems hijacks.

People · Andrew Ferguson

How this story unfolded

3 weeks · 6 reports · 4 community posts · from Sep 24

  1. Sep 24
  2. Sep 28
  3. Sep 30
  4. Oct 10
  5. Oct 12
  6. Oct 14
  7. Oct 15
  8. Oct 16

More stories today

Open the live feed