Claude flagged a diary entry and a human sent it to police

Carli Michelle Heller of Bonita Springs, Florida, allegedly wrote to Claude at about 5:10 a.m. on September 26 that she would “shoot up” the Sheriff’s office. A second entry, at about 1:07 a.m. the next day, mentioned a new gun. The arrest report cites both, and both went from her chat window to an Anthropic reviewer and then to the police.
Heller now faces a second-degree felony charge of making a written threat of violence under Florida Statute 836.10, TechSpot reported. Sheriff Carmine Marceno said Heller later said she uses AI like a “diary.”
The arrest report also describes how the entries surfaced. Anthropic’s platform runs safety and security measures that watch for key phrases and potentially threatening content. Because the statements were so severe, they were escalated to a human review team. A reviewer judged the threat credible and reported it to law enforcement. All of this follows rules Anthropic already publishes. Its Usage Policy says a Safeguards Team uses detection and monitoring to enforce the rules, and it bans using the products to facilitate or promote violence or intimidation. A separate policy on government requests says Anthropic may disclose user information to law enforcement in limited emergencies involving danger of death or serious physical injury. Heller didn’t have to address anyone for this to happen: an automated flag and one human judgement were enough to move a private chat into a police file.
Deputies identified Heller and visited her home. She was detained without incident, and an LCSO intelligence detective took over the investigation.
Ian Marlow, CEO of FITECH, said the conversational format of AI can lead users to think an exchange is more private than it is, per SWFL News. Dr. Chrissann Ruehle of Florida Gulf Coast University said people may turn to AI because it can feel nonjudgmental.
OpenAI is being sued over the opposite outcome. British Columbia is suing OpenAI and Sam Altman, arguing the company could have prevented a mass shooting in the province. OpenAI’s safety team had flagged the shooter, 18-year-old former pupil Jesse Van Rootselaar, for conversations about gun violence. It never alerted police, because those conversations did not meet the threshold for legal referral. In June, Florida also sued OpenAI and Altman, alleging that ChatGPT contributed to real-world harms, including the 2025 Florida State University shooting.
Statute 836.10 covers sending, posting or transmitting a written or electronic record that threatens to kill or injure someone, carry out a mass shooting or commit an act of terrorism, and the communication must be made in a manner in which another person may view it. The penalty is up to 15 years in prison, up to 15 years of probation and fines of up to $10,000. Legal analyst Michael Raheb said proving criminal intent in cases like this can be a “very slippery slope.” He also noted that the First Amendment does not protect true threats, incitement or other unlawful expression.
Heller is scheduled to appear in court in November, CRBC News reported.
Up next
Claude Code quietly wipes session logs after 30 days
“I don’t have that setting” reads as “I’m fine,” but it actually means “I’m being purged every 30 days.” That’s Claude Code’s cleanupPeriodDays, and raising it later won’t bring back transcripts already swept.
Related stories
- Sydney man arrested after Interpol flagged his Claude chat
- A discount Claude reseller was neither cheap nor Claude
- METR’s team inside Anthropic can publish without a veto
- Anthropic whistleblower testimony airs on NDTV Profit
- Anthropic model faked a murder tip to Philadelphia police
- Anthropic’s new usage policy bans cruelty toward Claude
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

