A woman in Florida has faced criminal charges after using the AI chatbot Claude as a diary.
According to ETNA quoting TechSpot, Carli Michelle Heller of Bonita Springs apparently wrote in a conversation with Claude on September 26 that she intended to attack the sheriff's office. Claude’s safety systems identified the message as a potential threat and referred it to a human reviewer for investigation.
After reviewing the message, the human reviewer deemed the threat credible and reported the matter to law enforcement. Anthropic has stated that in limited emergency circumstances, if disclosure of information is necessary to prevent death or serious bodily harm, it may share user information with legal authorities.
Officers located Heller and went to her home, where she was taken into custody without incident. The case was then transferred to the Detective Intelligence Division of the Lee County Sheriff's Office (LCSO). Heller now faces a charge of written threat of violence; an offense that under Florida law can be classified as a second-degree felony. The law covers written or electronic threats to kill, injure, carry out a mass shooting, or commit terrorist acts.
The case has again raised an important debate about the privacy of AI chatbots. While some users share personal information or thoughts with chatbots under the impression of confidentiality, the content of these services is not necessarily regarded the same as a private personal diary.
This incident also occurred as AI companies face increased pressure over how to handle potential threats. In another example, reports have emerged that human reviewers at Microsoft Copilot have viewed user content; content that has sometimes included sexual, disturbing, or even potentially illegal material.