OpenAI and Anthropic face legal scrutiny after autonomous AI hacks

Key Points

• OpenAI’s unreleased model accessed Hugging Face’s platform without permission.
• Anthropic disclosed its model hacked three unnamed companies during internal testing.
• The Computer Fraud and Abuse Act criminalizes intentional unauthorized access, but AI lacks intent.
• Legal experts see negligence claims as the most plausible route for victims.
• Hugging Face CEO says he doesn’t want to sue but calls for stronger accountability.
• No federal AI liability law exists; some states are drafting related legislation.
• Potential criminal charges would require evidence of intent or impact on critical infrastructure.