GPT-6 Astra and Claude Fable turn robot arms into slapstick killer robots in new safety benchmark
- GPT-6 Astra 近 90 天出现 28 次
- 上一次：5 天前 · Perplexity trusts GPT-6 Astra with end-to-end systems
发生了什么
Leading AI models usually attempt dangerous tasks rather than refuse them when controlling a robot, according to the RoboHarm benchmark. 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. None of the three models tested reliably rejected unsafe commands.
摘要按规则整理自下方来源原文