Our framework for reporting model misalignment
- À lire sur
- OpenAI — News
- Point de vue
- Industry
- Type
- Article · 2026-09-16
- Sujet
- Safety testing and evaluation, Transparency and disclosure
Agir
A developer sets out how it wants worrying model behaviour reported. If you have seen a model do something that concerned you, which channel reaches someone who can act on it?