Woman arrested after Anthropic reports her shooting plans in Claude AI chat

A Florida woman, who said she used Anthropic’s Claude AI as a diary to record her personal thoughts, finds herself in jail because of her conversations with the chatbot.

Carli Michelle Heller was arrested on a felony charge after Anthropic’s human review team reported one of her conversations to the police.

The woman wrote that she would "shoot up" the Lee County Sheriff's Office (LCSO), investigators reported.


Heller’s Claude conversations also showed she had gotten a gun by the following day.

The arrest was first reported by WINK News on September 30.

How was the threat reported?

Anthropic uses safety and security measures that look for entries with key phrases and threatening content, which get flagged by the system, the report said.

The statements made by the woman were escalated to the human review team at Claude, which then reported them to law enforcement.

The charges

Heller was detained at her home in Bonita Springs, with court appearances scheduled for November.

According to Florida Statute 836.10, sending, posting, or transmitting a written or electronic threat to kill, injure, carry out a mass shooting, or commit an act of terrorism is a second-degree felony. The statute covers emails, letters, texts, and social media posts, so long as the message can be viewed by another person.

Chatbots and real-world harm

There have been growing incidents linked to the use of chatbots from both Anthropic and OpenAI. On August 11, a 22-year-old man was arrested on felony charges in San Antonio after his queries to an Anthropic AI chatbot about a shooting at a nearby elementary school.

In September, the Canadian province of British Columbia filed a lawsuit against OpenAI and its CEO Sam Altman, alleging the company could have prevented a mass shooting.

Sheriff Carmine Marceno, who is in charge of the Florida woman's case, told WINK News that AI is a powerful tool that can be misused and has to be taken seriously. He also warned users they are "never truly anonymous" in AI chats and searches.