Do Son 
In late September, police in Lee County, Florida, arrested thirty-year-old Carli Michelle Heller. Court records show authorities charged Heller with violating Florida Statute 836.10. Specifically, she allegedly made written or electronic threats to conduct a mass shooting or act of terrorism. The incident originated when the AI developer monitored the chat logs and contacted the police.
The Escalation of AI Threats
Police cited the arrest report regarding this timeline. Heller wrote in a Claude conversation on September 26 that she intended to attack the Lee County Sheriff’s Office. Furthermore, she mentioned acquiring a new firearm the following day. The report stated that the developer’s security mechanisms actively monitor for specific keywords and potential threats. Due to their severity, the system escalated these particular messages to human review.
Law Enforcement Intervention
Subsequently, the review team reported the situation to law enforcement. Consequently, a Florida woman was arrested for allegedly making threats in an AI chat. Local authorities rushed to her residence and apprehended her immediately.
Using Claude as a Personal Diary
During a media interview, the Lee County Sheriff provided more context. He explained that Heller later claimed she used Claude as a personal diary. Essentially, she confessed all her private thoughts to the AI. However, the company’s privacy policy addresses this scenario clearly. It states that they may share personal data with government agencies or law enforcement.
Privacy Policy and Safety Protocols
They will do this if they possess a valid reason to believe the disclosure prevents severe harm to individuals or property. The privacy center also notes that employees typically cannot view user conversations. Yet, designated trust and safety team members can access user chats if required to enforce usage policies. Therefore, sharing this information with the police did not violate their new privacy agreement.
A Growing Trend in AI Monitoring
Moreover, this incident does not represent the first time the company proactively shared information with police. Since August, they have shared at least three Claude conversations with law enforcement. All these conversations involved relevant security incidents. However, details regarding the other two events remain undisclosed for now.
Support Our Threat Intelligence
Find our tech and OS security coverage helpful? Support our work today and unlock a 100% ad-free reading experience!
Buy Me a Coffee
