Anthropic Gave Police a Chat Log Containing Threats; Florida Woman Charged

Anthropic handed police the conversation between Florida resident Carli Michelle Heller and the Claude chatbot. According to an arrest report from the Lee County Sheriff's Office, on September 26, 2026, she wrote to the bot that she was going to carry out a shooting at the sheriff's office, and the next day said she had bought a new weapon. The woman was detained at her home in Bonita Springs and charged with written threats of violence; under Florida law this is a second-degree felony.

County Sheriff Carmine Marceno said Heller used AI "as a diary." In a statement to the press, he called artificial intelligence a powerful tool that can also be used to cause harm, and urged that threats created or transmitted with the help of AI be taken seriously. In his words, it is important for users to understand that in chats with AI, and in search queries, they do not remain truly anonymous.

At Anthropic, a unit called Safeguards monitors how people use Claude: it checks compliance with the service's rules and can slow down or cut off an offender's access, as well as adjust responses. The company did not immediately respond to a request for comment.

Not the first case

This is not the first time a conversation with Claude has been handed to police. In San Antonio, a conversation about a shooting at an elementary school led to the arrest of a 22-year-old: he was charged with third-degree terrorist threats.

Against the backdrop of this case, lawmakers are discussing how far AI companies themselves should be responsible for order on their platforms. Senators Josh Hawley and Chris Murphy are pushing a bill that could open the way for tech giants to face lawsuits.

Translation editor: Салтанов А.