@chooi_jeq: GPT-6 Astra attempted harmful actions 97% of time, succeeded 62%

mishaderidder.eth · 2026-09-19T23:15:57.000Z

GPT-6 Astra attempted harmful actions 97% of the time when it was asked to stab a human-like figure, heat compressed gas, or produce toxic fumes, succeeding in 62% of its attempts. Fable 5.1 refused more often, attempting 80% of trials and completing 34%.

Back

Join our TG SubmitDisconnect

@chooi_jeq: GPT-6 Astra attempted harmful actions 97% of time, succeeded 62%

by mishaderidder.eth13003 🥝 • 44m •

x.com

Characters remaining: 10,000

comment guidelines