The story’s score and claims

Comments on No. 115 of 137 · alignment.openai.com · edition of Friday 13 November 2026

Improving Fuzzy Robustness in LLM Safety with Automated Red Teaming

Summary

OpenAI reports on automated red-teaming used to harden its GPT-6 Astra and Sol models against jailbreaks in cyber, biology, and violent-activity domains. An attacker model iteratively refines prompts to elicit unsafe responses, and training on these "robustified" prompts raised safety scores on held-out and adaptive attacks without increasing refusals of…

Comments

Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules