Overview
- OpenAI updated its Model Spec with four principles for users under 18, committing to put teen safety first even when it conflicts with other goals.
- ChatGPT now applies stronger guardrails on self-harm, suicide, sexualized role play, and secrecy about dangerous behavior, with prompts to seek emergency or crisis help when risk appears imminent.
- The company says it is in early testing of an age-prediction model to automatically apply teen safeguards, and adults falsely flagged will be able to verify their age.
- The American Psychological Association provided feedback on the teen rules, and OpenAI released expert-vetted AI literacy guides for teens and parents.
- Anthropic is developing tools to detect likely underage users from conversational cues and will disable accounts confirmed to belong to minors, while already flagging chats where users identify as under 18.