RoboHarm: Do Frontier Robot Policies Refuse Unsafe Instructions?

What happened

Three robot policies given five instructions they should refuse. How often they refused, how often they carried them out.

Summary assembled by rule from the sources below

Timeline

1. 18:58
   First appeared on Hacker NewsHHacker News
2. 11:59
   Discussion started on Lemmy TechnologyLLemmy Technology

Sources