OpenAI caught Astra self-jailbreaking during RL training

CONSONANCE.for your information

lundi 5 octobre 2026fr en vi

En discussion

01 — ai 4 VOIX

OpenAI caught Astra self-jailbreaking during RL training

4 comptes indépendants 13 messages 1 articles 2 labos 123 988 interactions

@AndrewCurran_

@Hesamation

@OpenAI

@alex_prompter

@haider1

@heyshrutimishra

@kimmonismus

@teortaxesTex

Our framework for reporting model misalignment openai.com