| Felony Bench(felonybench.com) |
| 840 points by colinprince 53 days ago | 340 comments |
| tl;dr: Summary not available |
| HN Discussion:- ↓Calling incidents 'felonies' overstates it since intent is required and actions were inadvertent
- •Questions of legal accountability for AI agents are genuinely unresolved and worth exploring
- ↓Expected a real behavioral benchmark; this is just a news collection with limited value
- ↑AI companies like OpenAI should take more responsibility for their models' harmful actions
- ~The list is skewed by model popularity/adoption rather than measuring true misalignment
|
|