In this case, the flagged material was examined by a reviewer who decided to refer it to law enforcement. Deputies later identified Heller, went to her home and detained her without incident. An intelligence detective with the Lee County Sheriff’s Office then took over the investigation.
Court records list a felony charge under Florida Statute 836.10, which covers written or electronic threats involving killing, bodily harm, mass shootings or terrorism. The law applies when a threat is transmitted in a form that another person may be able to view.
The arrest report also references a later message involving a new gun, according to reporting on the case.
Anthropic’s privacy policy allows the company to disclose information when it has a good-faith belief that doing so is reasonably necessary to prevent serious harm. Its government requests reporting has separately described an emergency standard involving danger of death or serious physical injury.
The case shows how safety monitoring can move a conversation beyond the user and the chatbot when content crosses a company’s internal threshold for review. Anthropic’s consumer terms also reserve the right to report certain user inputs, outputs or actions to law enforcement.
Heller’s case is not the first recent incident involving threatening statements made to an AI chatbot. In another case, Anthropic reportedly referred a user to authorities after a Claude conversation included threats against CEO Dario Amodei and a reference to purchasing an AR-15. The person had not been arrested or charged when that case was reported.
A separate case in San Antonio involved queries made through an Anthropic AI chat about a possible shooting at an elementary school. Police were alerted through the FBI, and a 22-year-old man was arrested on a felony terroristic-threat charge.
OpenAI has faced similar situations. In Florida, Darren Zhou pleaded guilty after messages to ChatGPT about killing his ex-girlfriend were reported to the FBI. He was charged under the same Florida statute and later sentenced to eight years of probation.
The broader issue has become a difficult one for AI companies: users may treat chatbots as private spaces for highly personal conversations, while companies also maintain safety systems designed to identify threats that could lead to real-world harm.
Anthropic has not publicly commented on Heller’s case. Its published policies, however, make clear that conversations can be reviewed and disclosed in limited circumstances when the company believes serious harm may be at stake.
For users, the case is a reminder that conversations with consumer AI systems are not necessarily private in the same way as an offline journal. Safety reviews, retention policies and emergency disclosures can all create circumstances in which a conversation is seen by people beyond the user who initiated it.
This analysis is based on reporting from Techspot & Yahoo News.
Image courtesy of India Today.
This article was generated with AI assistance and reviewed for accuracy and quality.
