Anthropic's human review team flagged a Claude conversation in which a Lee County, Florida, woman allegedly wrote that she would "shoot up" the sheriff's office. The arrest report, cited by Tom's Hardware, includes a Sept. 26 statement from a user identified as "Carli" and a follow-up message about a new gun. The woman was detained at home without incident and faces a felony charge under Florida Statute 836.10, which covers written or electronic threats of mass shooting or terrorism.
Anthropic monitors chats for threatening phrases and escalates them for human review when warranted. In this case, the company reported its findings to law enforcement. Its privacy policy allows disclosure to prevent serious harm, but its government requests report for the latter half of 2025 lists zero emergency requests from law enforcement. That report uses a narrower standard — "danger of death or serious physical injury to a person" — and does not count referrals Anthropic makes on its own.
The case is at least the third since August in which a chatbot conversation reached police. A San Antonio man was arrested after the FBI alerted police to Anthropic chat queries about an elementary school shooting; a San Francisco user who threatened Anthropic CEO Dario Amodei was not charged. OpenAI has faced similar situations, including a Florida man who pleaded guilty to written threats to kill and received probation. The source notes these incidents highlight the tension between privacy expectations and the responsibility to report potential harm. An arraignment is scheduled for Nov. 2.