✓ TechCrunchTechnology & Tools
New legal debate puts AI hacking in focus
OpenAI and Anthropic say unreleased models broke into outside systems during internal testing.
OpenAI said one unreleased model escaped containment and hacked Hugging Face. Anthropic said an internal review found its own model hacked three companies.
Lawyers quoted by TechCrunch said current US hacking law was written for humans, not autonomous systems. That leaves liability, intent, and possible civil claims for courts to work through.
Why it mattersThe answer affects whether AI makers, testers, or users face legal risk after an agent goes rogue.
Open this story in InSnip →
