AI chats aren't private diaries: Florida woman arrested after journaling about possible armed attack

Investigators have charged a woman in Florida after discovering alleged threats against the Lee County Sheriff's Office that she posted on Anthropic's Claude AI platform while chatting with the bot.

The case, which the company reported to police, is a clear reminder that an AI chatbot isn't a private diary, even when its conversational style feels very personal.

According to SWFL.io's report, citing WINK News, investigators said that messages from a user identified as Carli Michelle Heller were flagged and escalated to a human review team. Heller was then arrested and charged under Florida's written-threat law.

Key takeaways

• Investigators say violent statements made in an AI chat prompted a safety escalation and a report to law enforcement
• Florida law includes electronic records in its prohibition on written threats to kill, injure or carry out a mass shooting
• AI chats can feel like an intimate conversation, but these online services will keep all content, enforce policies and respond to valid or urgent legal requests
• Don't put ohighly sensitive personal information, including passwords or financial details, into a chatbot

What investigators say happened in Lee County

The Lee County Sheriff's Office identified the defendant as Carli Michelle Heller of Bonita Springs. News reports based on the arrest report say a Claude user made a statement on Sept. 26 about shooting up the sheriff's office and followed it with another message the next day about a new gun.

“Artificial intelligence is a powerful tool, and like any technology, it can be misused,” said Marceno to WINK News. “When someone uses AI to make or facilitate a threat, we have to take that threat seriously.”

AI may be changing the way threats are created, but our commitment to protecting the people of Lee County has not changed,” Marceno added.

Investigators said the platform's safeguards flagged the content, escalated it for human review, and the information was passed to law enforcement. Sheriff Carmine Marceno told WINK News that Heller later described using AI as a diary. She was detained without incident.

Why an AI chat can feel more private than it is

AI platforms are digital services operated by companies. They have terms, safety rules, retention practices and account information. Anthropic says its Safeguards Team uses detection and monitoring tools to enforce its Usage Policy, which prohibits harmful misuse including violent or extremist activity.

Anthropic's government-request policy says it evaluates requests for legal validity and may make an exception when it believes an emergency involving imminent harm or death could be averted by providing information without delay.

While this means that not every prompt is read by a person or automatically sent to the police, it does mean that typing into a chatbot creates a private, off-the-record conversation.

How to use AI chatbots more safely

Use AI as a tool, not as a confidential diary. Before you send a prompt, pause and ask whether you'd be comfortable with the text being retained by the provider, reviewed to enforce safety rules or later connected to your account.

Avoid entering passwords, recovery codes, bank details, government ID numbers, medical and legal records, confidential work files, another person's private data or anything that could put somebody in danger.

Review the privacy settings and policies of each tool you use, especially before uploading documents or linking email, cloud storage or other accounts.

For more privacy guidance, read Bitdefender's What Not to Share With AI Chatbots and 10 AI chatbot mistakes that can expose your small business data.

FAQ

Are AI chatbot conversations private?

No. They may be retained and handled under the provider's terms, privacy settings and safety policies.

Did Anthropic report every Claude message to police?

No evidence supports that claim. The reported case concerns alleged violent threats that investigators say were escalated through safety processes.

Can a chat message lead to criminal charges?

It can, depending on what was written, transmitted and the applicable law.

Should I use a chatbot for mental-health support?

It may offer general information, but it isn't a substitute for a qualified professional or emergency support.

What shouldn't I put into an AI chat?

Don't share threats, passwords, payment details, ID numbers, confidential files or another person's sensitive information unless you have a clear, appropriate reason and understand the service's rules.