Florida Woman Arrested After AI Chatbot Flags Violent Threats Against Sheriff's Office

AI Chatbot Safety Systems Trigger Arrest

Carli Michelle Heller, a 30-year-old Bonita Springs resident, was arrested on September 30 after Anthropic Claude flagged threatening statements she made during conversations with AI chatbots1

1

2

2

. The AI chatbot's safety mechanisms detected violent language in messages dated September 26, where Heller allegedly stated she would "shoot up" the Lee County Sheriff's Office2

2

. Investigators noted that a follow-up message the next day mentioned acquiring a new gun2

2

.

AnthropicClaude monitors chats for key phrases and threatening content. When AI safety mechanisms flag potentially dangerous material, human reviewers examine the conversations before deciding whether to escalate to authorities2

2

. In this instance, the review team determined the threat warranted reporting to law enforcement, leading to Heller's detention at her home without incident2

2

.

Legal Consequences Under Florida Statute

Heller faces a felony charge under Florida Statute 836.10, which addresses written or electronic threats of mass shootings or acts of terrorism2

2

3

3

. The statute makes it a second-degree felony to send, post, or transmit threats to kill or injure someone in any manner that another person may view3

3

. Court records list the charge as "THREAT-TERRORISTIC-STATE OFFENSES-WRITTEN/ELEC THREAT MASS SHOOTING/TERRORISM ACT"2

2

.

Heller reportedly told investigators she used the AI chatbot as a "diary," believing her entries were private3

3

. However, the legal interpretation suggests that conversations with AI chatbots qualify as communications viewable by others—namely, the AI companies and their human reviewers2

2

. A public defender was appointed at the time of charging, with arraignment scheduled for November 22

2

.

Pattern of AI Companies Reporting Users

This marks at least the third such case reported to law enforcement since August2

2

. On August 11, a 22-year-old San Antonio man was arrested after making violent threats about shooting an elementary school in an Anthropic AI chat, with police alerted by the FBI2

2

. Days later, another user threatened Anthropic CEO Dario Amodei in an August 14 Claude conversation, claiming to have purchased an AR-15, though this individual was not arrested2

2

.

OpenAI has encountered similar situations. In March, the company reported user Darren Zhou to the FBI after he discussed killing his ex-girlfriend in ChatGPT messages2

2

. Zhou pleaded guilty in August to charges including written threats to kill under the same Florida statute and received eight years' probation2

2

.

User Privacy Versus Public Safety Debate

AI companies face mounting pressure to balance user privacy with public safety obligations. Anthropic's public privacy policy, effective September 10, states that disclosure to law enforcement may occur when the company has a "good-faith belief that disclosure is reasonably necessary to prevent serious harm to any person or to property"2

2

. Its government requests report for the latter half of 2025 lists zero emergency requests from law enforcement, though this doesn't include referrals Anthropic initiates independently2

2

.

The ethical scrutiny intensified after Florida's attorney general filed a lawsuit against OpenAI in June, citing the 2025 Florida State University shooting, and requested a temporary injunction last month2

2

. British Columbia also sued OpenAI last month for not warning police about the Tumbler Ridge shooter, whose ChatGPT account the company had flagged eight months before the attack2

2

.

What Users Need to Understand

Lee County Sheriff Carmine Marceno emphasized to WINK News that what users share with AI chatbots is "never truly anonymous"2

2

. Making violent threats in conversations with AI chatbots can result in human reviewers examining the content and reporting it to law enforcement1

1

2

2

.

Recent reports revealed that human contractors reviewing Microsoft Copilot's image editor can access users' prompts, uploaded photos, and AI-generated edits, including potentially disturbing or illegal material3

3

. Florida Gulf Coast University's Chrissann Ruehle, speaking to WINK News, called for caution when using AI products2

2

.

These incidents demonstrate that AI companies are expected to respect privacy while simultaneously being held accountable when they fail to report threats reported to law enforcement2

2

. The increasing scrutiny has prompted demands for stricter, independently approved guidelines for AI safety mechanisms and user accountability2

2

.