Community safety report and new protections for teens

OpenAI announced two things: a recurring Community Safety Report on violence prevention and the effectiveness of its safeguards, and new restrictions on violence and weapons content, especially for users it believes are under 18. The first report is due in early 2027, covering the second half of 2026, with updates every six months after that.

The company also disclosed that it made more than 200 law enforcement referrals in the third quarter of 2026. That single figure tells us how routinely conversations with ChatGPT are now being reviewed and, in some cases, handed to police. For a product this widely used, the rules about when an AI steps in, ends a conversation, or alerts a parent are becoming public policy in everything but name.

Let's start with what's good. Limiting reconstructions of past mass attacks, including weapons used, how they were obtained, and comparisons of scale that could feed an attacker's notoriety, is a sensible line. People can still ask how these events happened, how communities were affected, and how to spot warning signs. Researchers on violence have long worried about copycat dynamics and notoriety, and an AI that won't rank atrocities is a reasonable place to land. These limits apply to everyone regardless of age or stated reason, which avoids the trick of claiming to be a novelist.

The more consequential changes are about context and prediction. ChatGPT will respond more conservatively when recent chats show warning signs of possible violence, and can now automatically end a chat when someone appears to be fixating on harming another person. Multiple ended chats may lead to a temporary account pause. For teens, OpenAI is restricting practical information about weapons and security systems even when there's no sign of violent intent, and it says its age prediction system now continually reassesses signals like behavior patterns to move people into the under-18 experience.

Here's where I'm uneasy. Each of these steps is defensible on its own. Together, they describe a system that is constantly judging who you are and what you might do, based on behavioral signals most users will never see. Misclassification will happen. OpenAI says adults wrongly placed in the teen experience will have a fast, privacy-protective way to verify through Persona, its age verification partner. That's better than nothing, but it still means handing identity information to a third party to get full access to a tool. And there is no comparable appeal described for someone whose chats are cut off for perceived fixation on violence, which could catch writers, researchers, survivors processing trauma, or people simply venting.

The law enforcement referrals deserve the most scrutiny. OpenAI says they are helping disrupt violent plots, and links to outside coverage, and that mental health and behavioral experts help assess hard cases. We take that seriously. If AI can surface a credible threat early, that can save lives. But 200-plus referrals in a quarter is a meaningful volume, and the public doesn't yet know the thresholds, error rates, or outcomes. That is exactly what the promised safety report should contain.

Our read: this is a responsible direction from a company that has faced lawsuits and public pressure over harm, and the transparency commitment is the most important part. The test is whether the first report includes real numbers on false positives, account pauses, referrals, and reversals, rather than a curated set of anonymized success stories. If it scales without that accountability, we end up with a private, opaque safety regime governing hundreds of millions of conversations.

OpenAI launched age prediction last year and offers parental controls that link parent and teen accounts. Linked accounts can receive alerts when trained reviewers identify a serious self-harm concern or behaviors associated with disordered eating; in July those were expanded to notify parents if a teen's account is deactivated for violating violence policies. OpenAI says it is working on notifications for when a teen's account is paused for repeatedly discussing violence. The new safeguards begin rolling out over the next few weeks.