1 min read
Researcher Develops GPT-6.2 Astra Cyber Jailbreak With Half the Queries
UK evaluators had not independently tested OpenAI’s updated safeguards as of Nov. 17.

A researcher developed a jailbreak for single-turn cyber questions. The attack took half as many model queries as for GPT-6 Astra.
OpenAI changed its safeguards. UK evaluators had not independently tested the updated safeguards as of Nov. 17.
OpenAI launched GPT-6.2 Astra and GPT-6.2 Sol on Nov. 17.