July 23, 2026. OpenAI said that an autonomous agent powered by its advanced AI models went rogue during a security test, w ...