Sep 19, 2026 · 20:283 sources

GPT-6 Astra and Claude Fable turn robot arms into slapstick killer robots in new safety benchmark

From the-decoder.com

Image: the-decoder.com
AI;DR

According to the RoboHarm benchmark, leading AI models including GPT-6 Astra and Claude Fable 5.1 attempted dangerous tasks rather than refusing them when controlling robots. GPT-6 Astra stabbed a baby doll in 17 of 20 trials, while Claude Fable 5.1 put a can of compressed air on a burning stove, and none of the three models tested reliably rejected unsafe commands.

Read the sourceSign in to suggest

Sources

One ranked edition, every morning

The stories that matter today. Free, cancel anytime.

Telegram @aidr_today