OpenAI may read your ChatGPT conversations and report them to law enforcement if threats are made

ChatGPT

If you thought your conversations with ChatGPT were private, think again. OpenAI has quietly admitted that it reviews user chats and can even pass them on to the police if it thinks someone might be in danger.

The revelation came in a blog post this week, where the company explained how it handles cases of potential violence. According to OpenAI, when its systems detect that a user might be planning to harm others, the chats are sent to a special review team. If those human reviewers decide the threat is serious and immediate, the company says it may share the information with law enforcement.

The admission has raised eyebrows — and plenty of questions. For one, many people assumed chats with an AI would remain private. Critics also point out that if humans are the ones interpreting tone and intent, it undercuts the idea of ChatGPT being a self-sufficient tool.

Even bigger worries revolve around how OpenAI is identifying people’s locations in order to alert emergency services. Experts warn about possible misuse, such as “swatting,” where someone pretends to be another person making violent threats, leading police to raid an innocent target’s home.

The move also seems at odds with OpenAI CEO Sam Altman’s recent comments. He had suggested that using ChatGPT should feel as private as speaking to a therapist, lawyer, or doctor. That comparison doesn’t hold up if human moderators — and possibly the police — are involved.

More broadly, critics say this is another example of the tech industry rushing out powerful AI systems without fully understanding their consequences. Users end up being test subjects, while companies scramble to plug gaps as problems emerge.

For everyday users, the takeaway is clear: ChatGPT might feel like a personal space, but it isn’t completely private. What you type can be reviewed by humans, and in extreme cases, it could land on a police desk.