In Brief
In late September 2026, the Lee County Sheriff’s Office arrested a woman after Anthropic, the company behind the Claude AI assistant, reported her chat messages to police. According to the arrest report, she wrote that she would “shoot up” the sheriff’s office and, a day later, that she’d gotten a new gun. She’s charged under section 836.10, Florida’s written or electronic threats law, a second-degree felony. The case raises a question the statute’s words don’t settle: whether typing into an AI chat is sending, posting, or transmitting a threat “in any manner in which it may be viewed by another person.”
A lot of people type into an AI assistant the way they’d write in a notebook. The company on the other end doesn’t see it that way. Here, Anthropic’s systems flagged the messages, a human review team read them, and the team reported them to law enforcement, according to the arrest report as described by WINK News, which first reported the arrest on September 30.
She’s been charged, not convicted. Court records described by Tom’s Hardware show an arraignment, the hearing where a plea is entered, set for November 2.
On This Page
What Police Say Happened
WINK reported that, according to the arrest report, the woman “made a statement on Sept. 26 saying she was going to ‘shoot up’ the Lee County Sheriff’s Office,” and the next day made another statement “saying she had gotten a new gun.” The report says Anthropic’s platform uses safety and security measures “to monitor for key phrases and potentially threatening content,” and that because of the severity, “the information was escalated to a human review team, which then reported the statements to law enforcement.”
Deputies detained her at her home without incident. Sheriff Carmine Marceno told WINK that she later said she uses AI like a “diary,” and he said users should understand they are “never truly anonymous,” specifically when using AI chatrooms and searches. Tom’s Hardware reported no comment from Anthropic on the case.
What Section 836.10 Says
Section 836.10(2) makes it unlawful “for any person to send, post, or transmit, or procure the sending, posting, or transmission of, a writing or other record, including an electronic record, in any manner in which it may be viewed by another person, when in such writing or record the person makes a threat to: (a) Kill or to do bodily harm to another person; or (b) Conduct a mass shooting or an act of terrorism.”
It’s a second-degree felony, punishable by up to 15 years in prison and a $10,000 fine. In Dominguez v. State, No. 1D2024-1250, slip op. at 6 (Fla. 1st DCA Mar. 18, 2026), the First District Court of Appeal listed the four elements the State has to prove, drawing on Florida’s standard jury instruction: a threat in a writing or other record, that the person sent, posted, or transmitted it, that it was done “in a manner in which it may be viewed by another person,” and that the person “intended the threat to be a true threat.”
Is an AI Chat “Sent, Posted, or Transmitted”?
Until 2018, the law reached only a threat sent to the person threatened or to a family member. Amendments in 2018 and 2021 replaced that with today’s wording, which asks only whether the writing went out in a manner in which another person may view it.
A chat with an AI assistant doesn’t fit those words neatly. The text leaves the user’s device and lands on the company’s servers, which sounds like a transmission. Anthropic’s privacy policy says conversations can be “flagged for safety review,” and here a person at the company did read them. But the writer was talking to software, and an AI model isn’t “another person.” Whether the chance that an employee might read a chat meets that element is a real question, and the text doesn’t answer it. I haven’t found a Florida appellate opinion applying section 836.10 to a conversation with an AI assistant.
The closest case I’ve found is N.H. v. State, 358 So. 3d 477 (Fla. 5th DCA 2023). A student left a handwritten kill list on her classroom desk, where a teacher saw it. The majority found probable cause that leaving it in plain view in a public place was posting it. The dissenting judge reasoned that the statute punishes sending, transmitting, or posting a writing, not writing it, and that what turns a disturbing thought on paper into a threat is some act of communicating it. An AI chat sits somewhere between a notebook and a social media post, and that’s the disagreement a case like this may turn on.
A True Threat and a Mental State
The First Amendment protects jokes, venting, and exaggeration. It doesn’t protect a true threat. In Counterman v. Colorado, 600 U.S. 66, 74 (2023), the Supreme Court described true threats as “serious expression[s]” conveying that a speaker means to “commit an act of unlawful violence.” The Court also held that the government has to prove something about the speaker’s own state of mind, and that recklessness is enough. In the threats context, that means a speaker is aware “that others could regard his statements as” threatening violence and “delivers them anyway.” Id. at 79.
Florida’s appellate courts haven’t settled the exact standard under section 836.10. The Fourth District has said the defendant must have made the communication “with the knowledge that it will be viewed as a threat.” T.R.W. v. State, 363 So. 3d 1081, 1088 (Fla. 4th DCA 2023). In Dominguez, the First District didn’t decide the level of intent because no one had argued it in the trial court, and it held that the evidence was enough to send the intent question to the jury.
With an AI chat, that question gets pointed. Did the writer think anyone would read the words? Did the writer know flagged conversations get reviewed? A jury weighs those facts alongside the words themselves.
How a Company Reports a Chat Without a Warrant
The federal Stored Communications Act generally bars a company providing “an electronic communication service to the public” from knowingly disclosing the contents of stored communications. 18 U.S.C. § 2702(a)(1). One exception, § 2702(b)(8), lets a provider disclose contents “to a governmental entity, if the provider, in good faith, believes that an emergency involving danger of death or serious physical injury to any person requires disclosure without delay of communications relating to the emergency.” There’s no warrant, no subpoena, and no judge. It’s the company’s call.
The reports don’t say which legal path Anthropic used, and whether the act covers an AI chat service is a question of its own. Anthropic’s privacy policy, effective September 10, 2026, says it may share personal data with “government authorities, law enforcement, or other third parties” when it has “a good-faith belief that disclosure is reasonably necessary to” do several things, including “prevent serious harm to any person or to property.” I’ve covered the other ways police get account data in what a subpoena gets from an online account, and an earlier AI chat case in your AI chat is not a private journal.
Where a Case Like This Is Fought
First, the whole conversation. Two lines pulled from a chat aren’t the record. What came before and after, and how the assistant responded, bear on whether the words were a serious expression of intent. If the State relies on a later message about a gun, exactly what was said and when matters too.
Second, the transmission element. That’s a legal question about what the statute’s words cover, and it can be raised before trial.
Third, intent. Whether a message is a threat at all is usually for the jury. In Puy v. State, 294 So. 3d 930, 933 (Fla. 4th DCA 2020), a case under an earlier version of the statute, the court asked whether the message was “sufficient to cause alarm in reasonable persons.” The writer’s own state of mind is a separate question, usually proven with circumstances, because there’s rarely direct proof of what someone was thinking.
Fourth, authentication. Section 90.901, Florida Statutes, requires evidence “sufficient to support a finding that the matter in question is what its proponent claims” before an exhibit comes in. The State has to tie the account and the messages to the accused, usually through account records, login and device data, and a company witness. Who else could use the account, and whether the export is complete, are fair questions.
Anything typed into an AI tool sits on a company’s servers, can be reviewed under the company’s own rules, and can become evidence. My page on violent crime defense in Florida covers the related charges.
Common Questions
Can an AI company report your chats to the police?
Yes, in some circumstances. Anthropic’s privacy policy says it may share data with law enforcement when it believes in good faith that disclosure is reasonably necessary to prevent serious harm, and 18 U.S.C. § 2702(b)(8) lets a covered provider disclose contents to the government without a warrant in a good-faith emergency involving danger of death or serious physical injury.
Is a threat typed into an AI chat a crime in Florida?
Section 836.10 requires that a threat be sent, posted, or transmitted “in any manner in which it may be viewed by another person,” and that it be a true threat. Whether a private conversation with an AI assistant meets the first part is an open question as far as I’ve found.
Does the threat have to be sent to the person threatened?
No. That’s how the law worked before 2018. The current statute asks whether the writing was sent, posted, or transmitted in a manner in which another person may view it.
The Weekly Letter
Once a week I send a short letter about new Florida laws, trials worth following, and what they mean for you. Join the newsletter.
Further Reading
- Fla. Stat. 836.10, written or electronic threats, Online Sunshine
- 18 U.S.C. § 2702, voluntary disclosure of customer communications or records, Cornell Legal Information Institute
- Anthropic Privacy Policy, effective September 10, 2026
- Counterman v. Colorado, 600 U.S. 66 (2023), Cornell Legal Information Institute
- Dominguez v. State, No. 1D2024-1250 (Fla. 1st DCA Mar. 18, 2026), CourtListener
- N.H. v. State, 358 So. 3d 477 (Fla. 5th DCA 2023), CourtListener
- Woman arrested after AI threat against Lee County Sheriff’s Office, WINK News, September 30, 2026
- Florida woman arrested after Anthropic reported her Claude chat to police, The Next Web, October 5, 2026
WRITTEN AND REVIEWED BY
Rory Safir, ACS-CHAL Forensic Lawyer-Scientist
He is the author of The Florida DUI Trial Manual, an NHTSA-qualified instructor in the standardized field sobriety exercises, faculty of the National College for DUI Defense, and a former Assistant Public Defender in the Thirteenth Judicial Circuit.
This post is general information about Florida law, not legal advice, and it does not create an attorney-client relationship. The law discussed is section 836.10, Florida Statutes, read with sections 775.082, 775.083 and 90.901, Florida Statutes; 18 U.S.C. § 2702; Counterman v. Colorado, 600 U.S. 66 (2023); Dominguez v. State, No. 1D2024-1250 (Fla. 1st DCA Mar. 18, 2026), and the Florida standard jury instruction it describes; T.R.W. v. State, 363 So. 3d 1081 (Fla. 4th DCA 2023); N.H. v. State, 358 So. 3d 477 (Fla. 5th DCA 2023); Puy v. State, 294 So. 3d 930 (Fla. 4th DCA 2020); and the Anthropic Privacy Policy effective September 10, 2026. The facts of the arrest come from news reports by WINK News, The Next Web and Tom’s Hardware, and the person charged is presumed innocent. All as of October 6, 2026.
