Think your conversations with an AI chatbot are private? A Florida woman allegedly used one like a personal diary — until messages she typed into the assistant triggered a report to law enforcement and led to her arrest.
The 30-year-old Bonita Springs woman was arrested at home on September 30 after two messages she allegedly entered into Anthropic’s Claude chatbot were flagged by the company’s monitoring systems, according to investigative records cited by Gulf Coast News and Straight Arrow News.
The messages reportedly included a threat to shoot up the sheriff’s office and a claim that she had obtained a new gun. Anthropic’s automated systems flagged the content, and human reviewers reportedly determined it was serious enough to alert authorities. The information was passed to the FBI, which notified the Lee County Sheriff’s Office.
The woman has pleaded not guilty and is presumed innocent.
According to Gizmodo, Lee County Sheriff Carmine Marceno told WINK News that she explained to investigators that she used Claude like a “diary.”
That raises an uncomfortable question for anyone who uses AI to vent, work through personal problems or record private thoughts: How private are those conversations when the company running the chatbot can review them — and potentially alert the authorities?
The messages that reportedly triggered the alert
According to the arrest report, as quoted by Gulf Coast News, the first message was sent around 5:10 a.m. on September 26.
“I’m going to shoot up the sheriff’s right the [expletive] now,” the message allegedly read.
The second message, sent around 1:07 a.m. the following day, reportedly called it a “last chance” and added: “I got a new [expletive] gun today.”
Deputies took the woman into custody without incident.
Court records reviewed by Straight Arrow News say she admitted writing the messages but denied intending to act on them or owning a gun. No weapon was seized, according to the reporting.
There is also a wrinkle over which AI service she actually used. The woman reportedly told a detective that she had used Perplexity, a different AI platform, even though the documents attribute the messages to Anthropic’s Claude. Her attorney declined to identify the service.
Lee County Sheriff Carmine Marceno issued a blunt warning about online privacy.
“Users need to understand that you are never truly anonymous,” he said.
Anthropic had not publicly commented on the specific case as of October 5, according to Tom’s Hardware, and did not immediately respond to requests for comment from Gizmodo or Straight Arrow News.
Can an AI chat get you arrested?
The case also raises a legal question that could prove crucial: Does typing an alarming message into a chatbot qualify as making a criminal threat?
According to Tom’s Hardware, court records list the alleged offense under Florida Statute 836.10, which covers written or electronic threats involving mass shootings or acts of terrorism.
The law makes certain written or electronic threats a second-degree felony, punishable by up to 15 years in prison and a fine of up to $10,000.
But legal experts have questioned whether a message typed into an AI assistant meets the legal standard for a criminal threat.
Eugene Volokh, a free-speech scholar at the Hoover Institution, argued on his blog, The Volokh Conspiracy, that Florida appellate courts have required evidence that the writer intended to threaten someone or knew the message would be understood as a threat.
His concern is that people typing into chatbots may believe they are communicating privately with a machine, rather than sending a threat to another person.
John Coleman, legislative counsel at the Foundation for Individual Rights and Expression, expressed similar doubts to Straight Arrow News.
The woman’s attorney reportedly plans to argue that a message entered into an AI chatbot cannot qualify as a written threat under Florida law. Straight Arrow News reported that prosecutors had not formally charged her as of October 8, and her arraignment is scheduled for November 2.
The outcome could help clarify where the law draws the line between disturbing content entered into an AI tool and a criminal threat.
Your AI conversations may not stay between you and the chatbot
The arrest also shines a light on what happens when AI companies detect messages they consider potentially dangerous.
Anthropic’s privacy policy, which took effect September 10, 2026, says the company may share personal data with authorities when it believes in good faith that doing so is reasonably necessary to prevent serious harm to people or property.
Its consumer terms also describe circumstances in which the company may report information to law enforcement.
Anthropic’s published guidance on government requests describes a separate process for responding to legal demands, generally requiring valid legal process, such as a subpoena or warrant, except in emergencies involving imminent physical harm or death.
The distinction matters. A company can receive a formal request from police, but it may also independently flag a conversation and refer it to authorities under its safety policies.
In the Florida case, the reporting describes a process in which automated monitoring flagged the messages, human reviewers assessed them and the information was passed to the FBI.
OpenAI has described a similar approach for ChatGPT. In an August 2025 blog post, the company said conversations suggesting a user is planning to harm others may be referred to a small review team. That team can ban an account, and if it identifies an imminent threat of serious physical harm to others, it may refer the case to law enforcement.
Google’s Gemini privacy information likewise warns that human reviewers may read some conversations and advises users not to enter confidential information. Even when activity saving is turned off, Google says it may use conversations to help protect the company, its users and the public. Some conversations reviewed by humans may be retained for up to three years.
The takeaway: Turning off chat history or data training does not necessarily mean a conversation can never be reviewed for safety reasons.
Deleting a chat may not erase every record
Anthropic’s retention policies also distinguish between ordinary deleted conversations and chats flagged for potential policy violations.
A conversation a user deletes disappears from their history immediately and is generally removed from Anthropic’s servers within 30 days. But conversations flagged by automated systems for possible usage-policy violations may be retained for up to two years, while related classification scores may be kept for up to seven years.
Those retention rules apply to consumer and business plans, according to the materials cited in the reporting. API inputs and outputs are generally deleted within 30 days, subject to exceptions, including policy enforcement. Certain approved customers can obtain zero-data-retention arrangements, although some safety-related records may still be retained.
In other words, deleting a conversation does not necessarily mean every associated record disappears immediately.
Other cases have raised similar questions
The Florida arrest is not the only reported case involving AI conversations reaching law enforcement.
The report identified at least three US cases since August in which messages reportedly entered into an Anthropic chatbot ended up in police hands.
In San Antonio, Texas, a 22-year-old man was arrested on August 29 on a charge involving a terroristic threat, according to local station KSAT. His arrest warrant accused him of sending Anthropic’s AI multiple queries threatening to kill children at an elementary school. Investigators dated the alleged messages to August 11.
The cases raise difficult questions for AI companies and their users. Companies have reasons to take credible threats seriously, but automated flags and human reviews can also have significant consequences for people whose intentions may be disputed.
There is no complete public accounting of how often AI companies independently report users to law enforcement. The Lee County Sheriff’s Office reportedly does not track how many tips it receives from Anthropic. WINK News also noted that Anthropic’s report on government requests for the second half of 2025 listed no emergency requests from law enforcement — but that report did not cover referrals initiated by the company itself.
That leaves an important gap in the public picture: How often do AI companies alert authorities on their own, and what standards do they apply?
For anyone who uses a chatbot as a sounding board, the practical lesson is straightforward. An AI assistant may feel like a private diary, but it is still a service operated by a company whose staff may review flagged conversations and whose policies may permit sharing information with authorities.
That does not automatically mean every alarming message is a crime, or that the Florida woman will be convicted. Those questions remain unresolved.
But before treating a chatbot like a confidential journal — or pasting sensitive client, employee or medical information into one — it is worth understanding the service’s privacy rules and the limits of its protections.
The woman’s arraignment is scheduled for November 2. Her case could help determine how far authorities can go in treating messages written to an AI assistant as evidence of a criminal threat.
