Summary
- Carli Michelle Heller allegedly threatened the Lee County Sheriff's Office and mentioned acquiring a new firearm in subsequent messages.
- Anthropic's monitoring system flagged her posts, prompting a human review team to alert law enforcement.
- The company's terms allow them to report user content to authorities at their discretion.
A woman from Florida has found herself facing serious legal consequences after using an AI tool as a diary, with her writings being reported to the police by an Anthropic employee.
Carli Michelle Heller, residing in Bonita Springs, was taken into custody following revelations that on September 26, she had expressed intentions to “shoot up” the Lee County Sheriff's Office. According to local reports from WINK News, her Claude AI account reflected communications indicating she had acquired a new gun the following day.
The arrest report highlights that Anthropic’s platform is designed to identify potentially dangerous phrases and content. Due to the alarming nature of Heller's statements, they escalated the matter to a human review team, which subsequently notified law enforcement.
Law enforcement officials were able to locate Heller at her residence and arrested her without incident. She now faces charges for making a written threat of violence.
Anthropic’s policies are clear regarding user data. Their consumer terms state that they reserve the right to report user inputs, outputs, or actions to police at their discretion. Furthermore, their privacy policy, which will take effect on September 10, 2026, permits data sharing with law enforcement if there is a reasonable belief that such disclosure is necessary to avert serious harm.
As a result, even if users opt out of certain data handling practices, Anthropic retains the right to utilize their data for a period of three months instead of five years, implying that user conversations may never be fully confidential.
This situation is not an isolated incident. The San Francisco Standard previously reported that another user was referred to police by Anthropic after allegedly stating he had purchased an AR-15 and targeted CEO Dario Amodei. An employee conveyed that there was no immediate safety concern, choosing not to disclose the messages due to company policy.
The individual involved claimed to be joking, and he was not arrested or charged at the time of that report. An Anthropic spokesperson described the situation as an example of the company's safeguards functioning correctly, but did not clarify if threats made against their employees are treated differently.
This issue extends beyond Anthropic. Following a tragic school shooting in Tumbler Ridge, British Columbia, which resulted in eight deaths, OpenAI considered notifying Canadian authorities in June 2025 but ultimately decided that the user’s account did not indicate an imminent danger.
OpenAI’s updated August 2025 policy now routes threats to human reviewers who may escalate them to law enforcement as necessary.
Heller is scheduled to appear in court in November.