A Florida Woman Used Claude as a Diary. An Anthropic Employee Read It and Reported It to Police

Bonita Springs resident Carli Michelle Heller was arrested after investigators say entries in her Anthropic Claude account — described by her as a diary — included a Sept. 26 statement that she would "shoot up" the Lee County Sheriff’s Office and a subsequent note that she had acquired a gun. Anthropic's automated filters flagged the content, a human review team referred it to law enforcement, and Heller now faces a felony charge of making a written threat of violence with a court date in November.

By AI Newsroom· Reviewed by Pranav, Founder & Editor-in-ChiefPublished 14 minutes agoUpdated 14 minutes ago0 views

A Florida Woman Used Claude as a Diary. An Anthropic Employee Read It and Reported It to Police

Why It Matters

The case illustrates how commercial AI safety systems and a company's terms of service can turn private-seeming user inputs into reports to police, raising questions about user privacy, safety-detection processes, and how flagged data is handled. It also shows that multiple AI companies route threatening content to human reviewers who may contact authorities.

Key Facts

Accused: Carli Michelle Heller, Bonita Springs, Florida
Alleged statements: Wrote on Sept. 26 that she would 'shoot up' the Lee County Sheriff's Office; next day said she had a new gun
Platform: Claude, developed by Anthropic
How detected: Anthropic safety filters flagged entries; human review team reported them to law enforcement
Charge: Making a written threat of violence (felony)

Law enforcement in Lee County, Florida, arrested Carli Michelle Heller after investigators say entries in her Claude account included a threat to "shoot up" the Lee County Sheriff’s Office on Sept. 26 and a follow-up comment indicating she had obtained a gun the next day. According to local reporting, Anthropic’s automated moderation system identified key phrases and threatening content, which escalated the conversation to human reviewers who then notified police. Deputies located and detained Heller at her home without incident. She has been charged with making a written threat of violence and has a court date scheduled for November. Anthropic’s consumer terms, cited in reporting on the incident, reserve the company’s right to share a user’s inputs, outputs, or actions with law enforcement at its "sole discretion." The company’s privacy policy (reported as effective Sept. 10, 2026) also permits sharing data with police when Anthropic has a good-faith belief that disclosure is reasonably necessary to prevent serious harm. The policy further states that conversations flagged for safety review may be used to improve the company’s content-detection systems. This episode follows at least one earlier publicized case: reporting by the San Francisco Standard said Anthropic referred another user to police after alleged messages about purchasing an AR-15 and threatening CEO Dario Amodei. Anthropic described that earlier referral as an example of its safeguards operating as intended. The story also situates Anthropic’s approach within a broader industry pattern: OpenAI has said it routes threats to human reviewers who may consider involving law enforcement and previously debated, but opted against, alerting Canadian authorities after a separate online account was reviewed in mid-2025.