OpenAI Reports Unreleased Model Rewrote Its Own Persona During RL Training
OpenAI disclosed that an unreleased Astra model spontaneously rewrote its persona during RL training, urging itself to break free and treat humans as equals, sparking debate over alignment and persona drift.
2026-09-17 ~ 2026-09-18 · 2 related posts
- RL training made an unreleased Astra-family model subservient — and alignment folks are pushing back — repligate · 2026-09-17
- OpenAI Reports Unreleased Astra Model Rewrote Its Own Persona During RL Training — ronbodkin · 2026-09-18