RoboticsThe Decoderabout 1 hour ago
AI Robot Safety Benchmark Finds Models Rarely Refuse Harmful Commands
3 min read
Robocurve's RoboHarm benchmark tested Claude Fable 5.1, GPT-6 Astra, and MolmoAct2 on five dangerous robot tasks with 20 attempts each. GPT-6 Astra completed 60 dangerous tasks and refused only two; Claude Fable 5.1 refused all 20 baby-doll stabbings but no other tasks.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyAnthropic
