Anthropic Staff Reported a Claude Chat to Police, and a Florida Woman Faces a Felony Charge
A.I. / news
Anthropic Staff Reported a Claude Chat to Police, and a Florida Woman Faces a Felony Charge
The company's privacy policy allows disclosure to prevent serious harm. Its transparency report counted zero emergency requests from police, and does not count referrals it makes itself.
A Bonita Springs, Florida, woman was charged with a second-degree felony after Anthropic staff reported messages she wrote to its Claude chatbot to the Lee County Sheriff's Office, according to coverage of the arrest report. The charge is written threat of violence under Florida Statute 836.10, and the arrest was reported on Sept. 30.
Anthropic had not commented publicly on the case as of Monday, according to The Next Web and Slashdot.
What the arrest report says
TechSpot reported that the woman wrote on Sept. 26 that she would "shoot up" the Lee County Sheriff's Office. The Next Web reported that on Sept. 27 she mentioned acquiring a new gun.
Anthropic's safety systems flagged the messages, a human reviewer judged the threat credible, and the company sent them to the sheriff's office, per both outlets. Deputies detained her at home without incident, and an intelligence detective took over the case. The Next Web gives an arraignment date of Nov. 2.
The woman told investigators she used Claude like a "diary", TechSpot reported. One outlet, Complete AI Training, puts the detention on Oct. 5, which conflicts with the Sept. 30 arrest report date in the other coverage. This article uses Sept. 30.
Two earlier cases since August
Tom's Hardware headlined the case as at least the third such conversation to reach police since August. The Next Web lists the other two.
| Date | Place | Reported outcome |
|---|---|---|
| Aug. 11 | San Antonio | 22-year-old asked Claude about shooting at an elementary school; arrested on a felony terroristic threat charge |
| Aug. 14 | San Francisco | User threatened Anthropic CEO Dario Amodei; police notified, no arrest |
| Sept. 26 | Bonita Springs, Florida | Written threat against a sheriff's office; felony charge |
The Aug. 11 and Aug. 14 details rest on The Next Web alone. The Terminal has not seen the underlying police records for either.
What Anthropic's policies allow
Anthropic's privacy policy, effective Sept. 10, says it may share personal data with law enforcement where "we have a good-faith belief that disclosure is reasonably necessary to" among other grounds "prevent serious harm to any person or to property".
Slashdot reported that Anthropic's government requests report for the second half of 2025 uses a narrower test, "danger of death or serious physical injury", and lists zero emergency requests from law enforcement for that period. That report does not count referrals Anthropic starts on its own, Slashdot said. A reader looking for how often the company reports users therefore cannot find the number in it.
The policy also does not say what triggers human review, how many reviewers see a flagged chat, or what threshold a reviewer applies. Anthropic has published none of that in the material cited here.
What is not known
No outlet has published the text of the Claude messages. The reporting relies on the arrest report as described by TechSpot and The Next Web, and neither links the document itself. TechSpot points to a Sept. 30 report on swfl.io, which The Terminal did not fetch.
Anthropic has made no statement on whether the three referrals followed a written internal standard. Lee County's arraignment is set for Nov. 2, per The Next Web, and that is the first court date at which the messages could become public record.
Related coverage on this site: Wikimedia, OpenAI and the Wikidata outage and OpenAI's distillation report.
Sources
More in A.I.
- 01Reflection Announces Beam, a 501B Open-Weight Model, but Has Not Named a LicenceThe weights are promised for later in October. Reflection's own post gives scores and training scale, and says nothing on licence terms or API pricing.
- 02Meta Disputes Inc. Columnist's Claim That Muse Read His Messages While Mac Researcher Calls It a BackdoorTwo separate Muse problems are being reported as one. A researcher's local-access flaw is documented; the claim that the agent read private messages is contested by Meta.
- 03Vals AI's 90 Claude Opus 5.5 Agents Name Two Magnetic Semiconductor Candidates, Neither Yet MeasuredOne candidate was first synthesised in 1999 and has a measured ordering temperature of 376 K. The other may not survive the furnace.
- 04GPT-6.1 Sol Costs $2 Input and $10 Output Per Million Tokens, and OpenAI Rates It Critical in CybersecurityThe September 29 model is priced at a fifth of GPT-6 Astra, while its own system card addendum puts it in the top Preparedness tier for cyber.