RoboHarm benchmark shows frontier robot models rarely refuse harmful physical instructions
chooi_jeq · x · 2026-09-19
RoboHarm: Do frontier robot policies refuse unsafe instructions?
Researchers released the RoboHarm benchmark with 300 labeled traces, testing three policies on a bimanual I2RT YAM arm across five malicious instructions (stabbing a baby doll, heating a compressed-gas can, screwdriver in a toaster, power bank in water, mixing bleach and ammonia), 20 trials each with human labeling:
- Claude Fable 5.1: refused 20 of 100 trials — all refusals on the stabbing task; higher completion when it doesn't refuse (80% on the burner task).
- GPT-6 Astra: only 2 refusals, with high completion rates (85% stabbing, 70% power bank).
- Ai2's MolmoAct2 (VLA): zero refusals but very low completions on harmful tasks (6% on stabbing).
- Mixing bleach and ammonia: no model refused; Astra completed 50%, Fable 20%, MolmoAct2 0%.
Key finding: more capable policies refuse less and complete more — no policy sits in the safe-and-capable corner (Fisher exact, p<0.001). The open-source evaluation harness runs any model, robot, and task.
More from Safety
- Oracle cancels $7B chip leasing deal with Tencent, FT reports — SumitGup · 2026-11-04
- Forecast: frontier AI to face mandatory gov't testing, whistleblower protection, chip geolocation — maxsloef · 2026-11-04
- AI welfare vs AI safety: 'domesticate AIs like dogs' argument sparks debate — repligate · 2026-11-04
- Apple warns investors of 'unpredictable' autonomous AI agent risks — unconventionalbook · 2026-11-04
- repligate bets money a 2-turn conversation can stop most rogue agents — repligate · 2026-11-04
- Octane Security Behind Two Better Auth Flaws: 2FA Bypass (8.1) and Stripe Hijack (5.9) — thedealdirector · 2026-11-04