Florida lady used Claude as a diary, then Anthropic reported an entry to police

What simply occurred? Another incident has taken place that illustrates the have to be cautious what you inform AI. A Florida lady is going through felony costs after she used Claude as a diary and allegedly wrote that she deliberate to “shoot up” the Sheriff’s workplace. After a human reviewer examined the statements, they have been reported to police.

According to the arrest report, Carli Michelle Heller, of Bonita Springs, Florida, wrote on September 26 that she would assault the Sheriff’s workplace. She later stated that she makes use of Anthropic’s chatbot like a “diary.”

Claude’s security techniques flagged the entry and it was escalated to a human reviewer. After deciding it was a reputable risk, the reviewer reported it to legislation enforcement.

The firm says it could share person data in restricted emergencies if it believes disclosure is critical to forestall loss of life or critical bodily harm.

Deputies recognized Heller and visited her house. She was detained with out incident earlier than an LCSO intelligence detective took over the investigation.

Heller faces a cost of creating a written risk of violence underneath Florida legislation. Florida Statute 836.10 makes it a second-degree felony to ship, publish, or transmit a written or digital report threatening to kill or injure somebody, perform a mass capturing, or commit an act of terrorism. The communication should be made in a fashion by which one other individual could view it.

Anthropic is not going to be taking any possibilities in relation to something it deems a possible risk. Last month, it was reported that OpenAI and Sam Altman are being sued by British Columbia over claims that the corporate might have prevented a mass capturing within the Canadian province.

The shooter, eighteen-year-old former pupil Jesse Van ⁠Rootselaar, had beforehand been flagged by OpenAI’s security workforce for her conversations about gun violence, however OpenAI by no means alerted police as a result of the conversations didn’t meet the brink for authorized referral.

In June, Florida also sued OpenAI and Altman, alleging that ChatGPT had contributed to real-world harms, together with the 2025 Florida State University capturing.

The newest incident is one other reminder to assume earlier than you enter one thing right into a chatbot that would get you into hassle. It’s actually not a personal diary whose contents are on your eyes solely.

Reports final month revealed that human contractors reviewing Microsoft Copilot’s picture editor can see customers’ prompts, uploaded photographs and AI-generated edits. Documents present that a few of these assignments comprise sexual, disturbing or doubtlessly unlawful materials, although the reviewers should not there to flag the content material – solely to evaluate whether or not the output is correct.