Florida woman used Claude as diary, wrote ‘shoot up’ Sheriff’s office, Anthropic called police
By Anjali Sharma
WASHINGTON – Carli Michelle Heller, of Bonita Springs on Tuesday told deputies that she treats the chatbot like a diary.
Anthropic’s safety system flagged her messages from September 26 and 27, and a human team sent them to the Lee County Sheriff’s Office. She was detained at home.
A woman in Bonita Springs, Florida, has been arrested after Anthropic passed one of her Claude messages to police.
Investigators say she wrote that she would attack the Lee County Sheriff’s Office.
She said she uses the chatbot like a diary. She now faces a felony charge. Her court date is in November.
The Lee County Sheriff’s Office identified the woman as Carli Michelle Heller.
She lives in Bonita Springs. Sheriff Carmine Marceno has spoken publicly about the case.
An intelligence detective from the sheriff’s office took over the investigation after her arrest.
The arrest report describes two statements. On September 26, a user identified as Carli said she was going to “shoot up” the Lee County Sheriff’s Office. The next day, the same user said she had gotten a new gun.
TechSpot identifies that platform as Claude.
The report said Anthropic’s platform uses safety and security measures. These monitor for key phrases and threatening content. The system flagged the entries.
Investigators wrote that the statements were severe. Because of that, they were escalated to a human review team.
TechSpot reports that the reviewer judged the threat to be credible. The review team then reported the statements to law enforcement.
Deputies identified Heller from the report. They went to her home in Bonita Springs. She was detained without incident.
The sheriff told WINK News that Heller later said she uses AI like a diary. The intelligence detective then took over the case.
Heller is charged with making a written threat of violence under Florida law. TechSpot points to Florida Statute 836.10. Under that law, the offense is a second-degree felony.
The statute covers sending, posting, or transmitting a written or electronic record.
The record must threaten to kill or injure someone. It can also threaten a mass shooting or an act of terrorism. The message must be made in a way that another person may view it.
Marceno said AI is a powerful tool and can be misused. He said the office must take a threat seriously when someone uses AI to make or support it.
He told users they are “never truly anonymous” in AI chats and searches.
He said AI may change how threats are made. He said the office’s duty to protect Lee County residents has not changed.
Anthropic’s Usage Policy says its Safeguards Team uses detection and monitoring to enforce the rules. The policy bans using its products to promote or support violence or intimidation.
The company also has a government request policy. It says user information may go to law enforcement in limited emergencies. Those cases involve danger of death or serious physical injury.
British Columbia sued OpenAI and its chief executive, Sam Altman. The province claims the company could have prevented a mass shooting there.
The shooter was Jesse Van Rootselaar, an 18-year-old former pupil. OpenAI’s safety team had flagged her conversations about gun violence.
OpenAI did not alert police. The company said the conversations did not meet the threshold for legal referral.
In June, Florida also sued OpenAI and Altman.
The state alleges that ChatGPT contributed to real-world harm. That includes the 2025 shooting at Florida State University.
In a separate report from last month covered Microsoft Copilot.
It said human contractors who review the image editor can see user prompts. They can also see uploaded photos and the edits the AI makes.
Some of that material is sexual, disturbing, or possibly illegal. The reviewers are not there to flag it. Their task is to judge whether the output is accurate.
