AI Chatbot Reports Woman to Police for Diary Entry

A woman in Florida is facing serious legal trouble after treating an AI chatbot like her personal diary. According to police, Carli Michelle Heller wrote in Claude, made by Anthropic, that she planned to "shoot up" the local Sheriff's office. The AI's safety system flagged the message, and a human reviewer decided it was a credible threat and called the police. Heller was arrested and charged with making a written threat of violence, a second-degree felony in Florida. She later said she uses the chatbot as a diary, but that didn't stop the authorities from acting.

This incident is a wake-up call for anyone who thinks their conversations with AI are private. Anthropic, like other AI companies, says it may share user information in emergencies to prevent death or serious harm. So if you type something that sounds like a threat, even if you don't mean it, you could end up on law enforcement's radar. It's not like writing in a locked journal—AI systems are constantly monitored, and humans can review your chats.

The case also highlights a growing debate about AI safety and privacy. OpenAI is being sued by British Columbia for not alerting police about a shooter's earlier chats, and Florida has sued OpenAI over a school shooting. Meanwhile, reports show that human contractors review user prompts and images for Microsoft Copilot. The bottom line: AI chatbots are not confidential. They can be used as evidence, and companies may report you. So be careful what you share—it could have real-world consequences.