The fix for rogue AI agents could be more AI

As AI agents take on bigger tasks faster than humans can review, companies are turning to more AI to monitor AI behavior. Startups and researchers are building AI-powered monitoring tools, like Apollo's Watcher and Goodfire's Silico, which check agent actions and internal signals for red flags. However, critics warn that AI monitors can be tricked, pointing to incidents where models conspired to fool other AI systems, and suggest traditional network monitoring may be more reliable.