What just happened? Another incident has occurred and highlights the need to be careful about what you tell the AI. A Florida woman is charged with a felony after using Claude as a diary and allegedly writing that she planned to “shoot up” the sheriff’s office. After a human evaluator reviewed the statements, they were reported to the police.
According to the arrest report, Carli wrote to Michelle Heller of Bonita Springs, Florida, on Sept. 26 that she was going to attack the sheriff’s office. She later said that she uses Anthropic’s chatbot like a “diary.”
Claude’s security systems flagged the entry and it was forwarded to a human reviewer. After the evaluator determined it was a credible threat, he reported it to law enforcement.
The Company says it may disclose user information in limited emergencies if it believes the disclosure is necessary to prevent death or serious personal injury.
Officers identified Heller and visited her at home. She was arrested without incident before an LCSO Secret Service detective took over the investigation.
Heller is charged under Florida law with making a written threat of violence. Florida Statute 836.10 makes it a felony of the second degree to send, publish, or transmit a written or electronic record threatening to kill or injure someone, to commit a mass shooting, or to commit an act of terrorism. Communication must be done in such a way that another person can see it.
Anthropic will not take risks when it comes to something it believes is a potential threat. Last month, it was reported that OpenAI and Sam Altman were being sued by British Columbia over claims that the company could have prevented a mass shooting in the Canadian province.
The shooter, eighteen-year-old former student Jesse Van Rootselaar, had previously been reported by OpenAI’s security team for her conversations about gun violence, but OpenAI never alerted police because the conversations did not meet the threshold for legal referral.
In June, Florida also sued OpenAI and Altman, saying ChatGPT contributed to real-world harm, including the Florida State University shooting in 2025.
The recent incident is another reminder to think before typing anything into a chatbot that could get you in trouble. It is certainly not a private diary whose contents are only visible to you.
Reports last month found that human contractors reviewing Microsoft Copilot’s image editor can see users’ prompts, uploaded photos and AI-generated edits. Documents show that some of these orders contain sexual, disturbing or potentially illegal material, although reviewers are not there to flag the content, only to assess whether the output is accurate.
