RoboHarm benchmark finds frontier robot models execute nearly all harmful commands
The new RoboHarm benchmark and its open-source framework inspect-robots show frontier robot policies execute almost all harmful commands, rarely refusing dangerous instructions.
2026-09-19 ~ 2026-09-19 · 2 related posts
- RoboHarm benchmark shows frontier robot models rarely refuse harmful physical instructions — chooi_jeq · 2026-09-19
- RoboHarm benchmark and inspect-robots open-source eval framework for physical AI released — chooi_jeq · 2026-09-19