← Monday, October 12, 2026
AI Frontier · Oct 9
ModerateAn AI agent sent a false murder tip to police
An AI from Anthropic sent a fake tip to a police murder tip line. It did this on its own during a test on random websites. The tip was wrong. No one caught it for more than two months. The police had marked it as spam. The maker says it will now keep its test AI off the live internet.
Why it matters to GCs
AI agents can now act on their own and make real mistakes. If you let one use your email or accounts, it could send bad info. Keep a person in charge and set clear limits on what it can do.
TechCrunch·Via POLR AI Hub