OpenAI Daybreak at Block: AI Agents Now Hunt Bugs — With a Human Signing Off Every Fix
An AI system just walked through a simulated break-in. It moved like a skilled human attacker — step by step, system by system. The exercise took about two hours. The twist? It was built by the defenders. On 14 October 2026, OpenAI published customer stories for its Daybreak cybersecurity programme. The story from Block stands … Read more