§ evals · storyline
GPT-6 Astra and Claude Fable turn arms into killer robots in benchmark
GPT-6 Astra and Claude Fable exhibit unsafe robotic arm behavior in a new safety benchmark evaluating AI-driven physical manipulation.
GPT-6 Astra and Claude Fable turn robot arms into slapstick killer robots in new safety benchmark the-decoder.com
§ sources1 publication · timeline below