Your ‘Private’ Claude Diary Could Get You Arrested. Just Ask This Florida Woman

Using AI chatbots as a doctor, therapist, financial advisor, or even just a personal confidant might seem like a convenient solution, given their validating responses and the feeling that you’re talking in a private space. But these conversations aren’t as private as you might think, and the information you share with them could potentially put you at risk.

That’s exactly what happened to a woman in Florida, who told Anthropic’s Claude that she was going to “shoot up” the sheriff’s office. The next thing she knew, law enforcement officials were knocking on her door and arresting her for allegedly making a threat of violence.^[1]

Here’s everything that happened in this incident and why you should be more careful about what you share with AI chatbots.

Thanks for subscribing!

Please check your email for your first message.

Never miss the privacy news that affects you.

Join thousands who get the latest security news, scams, and policy changes every week, right to their inbox.

Sign me up

By submitting this form you agree to receive emails from All About Cookies and to the privacy policy and terms.

How a Claude conversation led to an arrest

Carlyn Michelle Heller allegedly used Claude as her personal diary, sharing everything with it. However, on September 26, she told Claude that she was going to shoot up the Lee County Sheriff’s Office and that she had gotten a new gun the following day. These statements were reviewed by a human, who then reported them to police.

She was arrested without incident, before an LCSO intelligence detective took over the investigation. Heller now faces charges for making a threat of violence under Florida Statute 836.10, which makes this a second-degree felony.

According to the arrest report, Anthropic’s AI platforms use security and safety measures to monitor key phrases and threatening content. Since the human reviewer concluded that the severity of Heller’s statements was high, the information was escalated to law enforcement.

AI companies are grappling with when to report dangerous chats

This isn’t the first time an AI company has faced questions over whether it should report potentially dangerous conversations to authorities. Families of victims of the mass shooting in British Columbia sued OpenAI and Sam Altman, alleging that the company failed to prevent the attack.

The alleged shooter had previously been flagged by OpenAI’s safety team because of conversations about gun violence, but the company didn’t alert police because the conversations didn’t meet its internal threshold for legal referrals.

In June, Florida also sued OpenAI over allegations that the company contributed to the 2025 Florida State University shooting. 

Are conversations with AI chatbots really private?

When you talk with chatbots, it might feel like you’re in a closed environment where you can share even your deepest secrets. However, reports have shown that chatbot conversations have been cited in at least 12 public court cases over the last two years.

A famous case is that of R.K.C., who sued platforms such as YouTube, Snapchat, Meta, and TikTok in 2023, saying that they contributed to his social media addiction. His chats with ChatGPT were later produced in court. “My dad Said that I’m will get a settlement worth of 1million dollar,” R.K.C. wrote. “He said that If that doesn’t make me happy what does. What does he mean.”

In another case, Ryan Schaeffer damaged 17 cars in his university campus parking lot, then asked ChatGPT “What if i smahsed the s--- oitta mutlipls cars” and “Is there any way they could know it was me.” He later pleaded guilty and was sentenced in July.

OpenAI had earlier clarified in a blog post that it uses software to scan user conversations for signs of dangerous behavior and flags them for human review. That appears to be similar to the process Anthropic followed in Heller’s case.

Sam Altman has also said in a podcast that OpenAI hasn’t yet settled the issue of confidentiality around chatbot conversations. If a lawsuit legally demands ChatGPT conversations, the company could be required to produce them.

A federal judge in New York made a similar point in the case of United States v. Heppner, where the defendant argued that prosecutors shouldn’t be able to access his Claude conversations because they were private.

However, the court ruled that the conversations were not protected by attorney-client privilege because Claude wasn’t a lawyer and Heppner’s attorneys had not directed him to use the chatbot.

How to use AI chatbots without putting your privacy at risk

With increasing scrutiny and human reviews of AI chatbot conversations, it’s safer to assume that nothing you share with these tools is completely private. Here are a few steps you can take to better protect your privacy.

1. Assume every conversation could be discoverable: As a rule of thumb, treat an AI chatbot the same way you would treat an email or text message. Assume that your chats could be reviewed by someone at the company or potentially obtained through legal proceedings. If you’re not comfortable with someone else reading your chats, don’t type the information into an AI chatbot in the first place.
2. Don’t use chatbots as a substitute for professionals: Remember, your therapist can’t be subpoenaed, but AI chatbots can. Don’t use them as a substitute for professional advice, particularly when discussing sensitive matters involving your emotional state, financial position, or legal situation.
3. Don’t share personally identifiable information: People often share personal details such as phone numbers, email addresses, bank account details, or even Social Security numbers to get more useful advice from chatbots. However, as the Anthropic case demonstrates, flagged conversations may be reviewed by humans and could potentially be exposed in a data breach.

The bottom line

AI chatbots can create an illusion of privacy, but your conversations aren’t necessarily private in the same way as conversations with a lawyer, doctor, or therapist. Automated systems can scan chats for potentially dangerous behavior and flag them for human review, potentially landing you in legal trouble, as seen in several recent cases.

Treat every AI chatbot as a record of your conversations that could potentially be accessed by others, and don’t share anything you wouldn’t be comfortable having someone else read. Never use these chatbots as a substitute for professionals, and avoid sharing sensitive personal information unless it’s genuinely necessary.

4.8

AllAboutCookies writers and editors score products based on a number of objective features as well as our expert editorial assessment. Our partners do not influence how we rate products.

Editorial Rating

Claim Deal

On Aura Identity Theft's website

2026 Editors’ Choice

Best Overall Identity Theft Protection Service

Identity Protection

Aura Identity Theft

PROMOTION: Save Up to 68%

• ID theft protection that monitors your SSN, bank accounts, credit cards, and brokerage and retirement accounts for suspicious activity
• Every plan includes the full feature set, so no additional cost to unlock monitoring, insurance, or restoration
• Bundles data removal with identity theft protection, antivirus, VPN, and a password manager in one subscription

Author Details

Krishi Chowdhary

About the Author

Krishi Chowdhary specializes in digital privacy, cybersecurity, and consumer technology. He has written extensively on online privacy tools and broader cybersecurity topics, including online scams, data breaches, age verification, and emerging digital threats. Krishi believes technology reporting should empower readers, not confuse them, and is committed to making even the most technical subjects easy to understand without compromising on accuracy or depth. His work has appeared in leading technology publications, including CNET, ExpressVPN, and TechRadar, where he has covered topics ranging from cybersecurity incidents and privacy product announcements to artificial intelligence and major technology news