Advancing red teaming with people and AI

In three linesOpenAI publishes an approach combining humans and AI for red teaming (adversarial security testing). The method improves vulnerability detection by leveraging the respective strengths of human testers and AI models to identify security flaws.

Summary generated by Claude — human-verified