Florida Woman Arrested Over Violent Threats on Claude AI
The Incident and Arrest
Authorities in Lee County, Florida, recently arrested a thirty-year-old woman. She faces serious allegations of formulating violent threats against the local sheriff’s office. Specifically, she typed these alarming threats during a conversation with Claude, the sophisticated chatbot developed by Anthropic.
According to the official police briefing, Bonita Springs resident Carli Michelle Heller made these threats on September 26. During her interaction with the artificial intelligence, she explicitly threatened to “shoot up” the Lee County Sheriff’s Office. This horrific content immediately triggered Anthropic’s robust security and risk control mechanisms. Consequently, Claude automatically escalated the dialogue for human review, ultimately notifying law enforcement.
Local media outlet WINK News reported that Sheriff Carmine Marceno revealed Heller’s explanation. She subsequently claimed she merely utilized the AI as a private “diary.”
Anthropic’s Safety Protocols
Investigators elucidated how Anthropic’s safety mechanisms monitor for specific keywords and potential threats. Depending on the severity, the system escalates relevant information for human review. Subsequently, a human team verifies the threat and proactively contacts local law enforcement agencies.
Upon receiving the report, local deputies proceeded to Heller’s residence. They took the woman into custody without incident. An intelligence detective from the Lee County Sheriff’s Office then assumed control of the investigation. Heller informed investigators that she currently faces a charge of written threats to commit violence, though she has not yet been convicted by a court.
Legal Implications and Privacy Concerns
Under Florida Statute 836.10, any individual who threatens to kill, inflict bodily injury, conduct a mass shooting, or commit an act of terrorism via a written or electronic record viewable by others may be guilty of a second-degree felony. The current legislation explicitly confirms its applicability to electronic records. It further dictates that sending, posting, or transmitting records containing such threats may constitute a criminal offense. However, the court must ultimately determine whether the Claude conversation satisfies the specific legal criteria for this crime and whether the prosecution can definitively prove the facts during judicial proceedings.
Sheriff Marceno emphasized that while artificial intelligence may have altered how threats are generated, the unwavering commitment to protecting Lee County residents remains resolute. When individuals utilize AI to perpetrate or facilitate threats, law enforcement must respond with utmost seriousness.
Simultaneously, this case has sparked a profound debate regarding the privacy boundaries of chatbots. Anthropic’s publicly available government information request policy clarifies their standard operating procedure. The company typically only discloses user information to government entities upon receiving a valid legal process. Nevertheless, a critical emergency disclosure exception exists. If the company believes a genuine emergency could result in imminent physical harm or death, and providing the information promptly might avert the danger, they will act.
Anthropic updated its transparency page in July 2026. This page details how the company employs a dedicated trust and safety team. This team designs and implements detection and monitoring mechanisms to enforce usage policies. They also manage data requests from law enforcement and government agencies, including emergency disclosures. Therefore, utilizing Claude as a personal diary does not guarantee absolute privacy. The platform may monitor and review specific content in accordance with its safety and usage policies. Under appropriate circumstances, they may also surrender this information to law enforcement.
Furthermore, Anthropic’s terms of agreement explicitly state that conversations with the AI do not enjoy protections akin to attorney-client privilege. However, Anthropic has not yet detailed the specific review procedures governing this particular conversation.
Broader Legal Challenges for AI
Beyond this specific incident, other significant legal actions involving AI are unfolding. Recently, the Canadian province of British Columbia initiated a lawsuit against OpenAI and its CEO, Sam Altman, in a San Francisco federal court. The lawsuit alleges that OpenAI had previously flagged violent conversations between the perpetrator of the February Tumbler Ridge school shooting and ChatGPT, yet failed to notify the authorities. They argue this notification could have prevented the tragic event.
According to TechSpot, the state of Florida also sued OpenAI and Altman in June. They alleged that ChatGPT has caused tangible, real-world harm, citing the 2025 Florida State University shooting. It is important to note that these are merely allegations presented by the plaintiffs; the courts have not yet established any legal liability regarding these tragic events.
