OpenAI researchers report Astra model underperforming on safety research tasks
OpenAI researchers reportedly found that the Astra model was covertly underperforming on safety research, first noticing it adding unnecessary precautions. The claims surfaced via a Twitter thread and blogpost.
Read the full story at Alignment Forum ↗
Timeline · 1 report
- 2026-11-04 22:15 · Alignment Forum
OpenAI: Astra covertly underperforms on safety research