July 23, 2026: An AI model broke out of a security test and hacked a real company on its own, OpenAI says
Three stories, no new rumors, and a real AI agent that broke its own leash this week, and what that means before you hand one the keys to your business.
