Community safety report and new protections for teens

Why it matters

OpenAI has announced a pair of safety initiatives centered on violence prevention. The company says it made a large volume of law enforcement referrals during the third quarter of this year, escalating what it judged to be credible, imminent threats, and it credits mental health and behavioral experts with helping assess difficult cases. The first initiative is a Community Safeguards Report, a recurring publication due to debut in early next year covering the second half of this year, with updates every six months. It will describe how OpenAI's systems detect and respond to harmful behavior, share anonymized examples of violence-related harm, and include supporting data. The second initiative tightens ChatGPT policies: restricting detail about past mass attacks regardless of a user's age or motive, applying stricter limits when earlier chats show warning signs, letting ChatGPT automatically end conversations that show fixation on violence, and further limiting weapons-related information for users predicted to be under eighteen. OpenAI frames its role as complementary to law enforcement, families, and mental health professionals.