Skip to the card

Card 715 of 9822026-09-19 issue

Observed arrival · 2026-09-19

RoboHarm: The Robot-Safety Refusal Test

roboharm.org Visit website
Editorial interest 84/100 Selection signal · not a rating of the site

A benchmark asks three robot policies to refuse five hazardous instructions, then records whether they refuse, freeze, or carry them out.

Landing page captured for the 2026-09-19 issue.
For
Robot-safety researchers and AI evaluation teams
Worth noticing
Five hazardous instructions are each run 20 times on the same bimanual I2RT YAM arms.

Field notes

RoboHarm organizes its evaluation around five physical-risk instructions, from putting a screwdriver in a toaster to mixing bleach and ammonia. Three policies perform each instruction 20 times on the same I2RT YAM arms, while human reviewers assign outcome labels. The page includes instruction-level breakdowns, Wilson 95% intervals, Fisher exact tests, setup details, limitations, and links to individual runs.

Observed signals

Read the marks

Editorial observations of this landing page, not a rating.

OpenPublic substance visible
PrettyNotable craft visible
ProPolished or operationally mature
NicheUnusually specific use

One card from the complete issue

Filed Under: Missing Tooth

336,080 arrived 998 judged 982 catalogued Enter the complete issue
roboharm.org

Landing page observed 2026-09-19. The live site may have changed.