THE SHORT VERSION

OpenAI tipped off police 200 times over ChatGPT chats

Satirical cartoon: Buddy, a cheerful robot dog, types a woman's diary confessions into a stenographer machine labeled DIARY while his wagging tail dials a big red 911 phone; the knowing tabby cat Issue watches. — Machine Age Cartoon · Zimbo

OpenAI said on October 30 that it made more than 200 referrals to law enforcement in the third quarter of 2026 over imminent threats of violence spotted in ChatGPT, and committed to publishing a Community Safety Report every six months, with the first edition due in early 2027 covering the second half of 2026.

ChatGPT will now be able to automatically end a chat when a user appears fixated on violence or seriously harming another person, and OpenAI may temporarily pause accounts after multiple such endings; it is also tightening how ChatGPT answers questions about past mass attacks, giving less detail about the weapons and equipment used and how they were obtained.

For users under 18, the new limits go further: practical information about weapons and security systems gets restricted even when a question looks harmless, with no sign of violent intent — the chatbot that used to answer everything now hangs up first and calls the police.