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

Leading AI models usually attempt dangerous tasks rather than refuse them when controlling a robot, according to the RoboHarm benchmark.