OpenAI launches ChatGPT for Teens with new safety defaults

This launch marks a deliberate step by OpenAI to define how its AI should behave for a younger audience, distinguishing between classroom use and independent study. The company frames ChatGPT for Teens as a support tool for moments when help is unavailable, using design choices grounded in learning science, such as guiding questions and self-explanation, rather than simply delivering answers. This separates it from the ChatGPT for Teachers offering, which remains institution-managed and classroom-focused.

The timing builds on a series of earlier commitments, including parental controls, an updated Under-18 Model Spec, and a Teen Safety Blueprint. The new version adds practical layers to those commitments: protections are on by default for under-18 users, with age-appropriate safeguards in high-risk areas like self-harm and eating disorders. Notably, the model spec now goes beyond blocking romantic roleplay to also prohibit romantic language and implying consciousness, reinforcing the goal of healthy real-world relationships.

The announced partnership with CodeAI, focused on education around AI rather than the technology alone, suggests OpenAI sees teens as a generation that should learn to direct and question AI. The stakes for this rollout likely hinge on whether the default protections can reduce risks without stifling the kind of creative, community-focused projects highlighted in the announcement, such as WiFind and Audemy. If the safeguards prove too restrictive, teens may simply seek AI tools elsewhere, which would undercut the purpose of the launch.