OpenAI introduced a ChatGPT experience for users ages 13 to 17 with stricter defaults and homework-oriented guidance. The product adds stronger content restrictions and guided study features, while age assurance and long-term memory remain points of criticism. ChatGPT for Teens applies stronger restrictions around self-harm, sexual and romantic content and other age-sensitive material.
Its homework support is designed to guide students through questions rather than immediately supply a final answer. The service does not require universal identity-based age verification; age assurance is used to estimate whether an account belongs to a minor. The launch can be verified at the feature and policy level, but its effectiveness cannot be inferred from design intent. Restrictions must be tested against ordinary use, adversarial prompts and ambiguous signals of distress.
Parental controls are opt-in and can manage quiet hours, voice, images, memory and some safety notifications. Critics told AP that long-term memory should be more restricted by default for teens. Common Sense Media research cited by AP found more than 70% of teens had used AI chatbots for companionship and about half did so regularly. The age boundary is equally important. If assurance misses a minor, protections may not apply; if it misclassifies an adult, the system may impose restrictions and require additional proof. Public error data would make that tradeoff measurable.
OpenAI previously published under-18 behavior principles, parental controls and an age-prediction approach. Age assurance trades stronger child protections against possible errors and privacy concerns for adults and minors.
Product defaults matter because many families will not change advanced settings. The current evidentiary limit is that independent evidence about error rates, harmful-content bypasses, learning outcomes and adoption was not yet available for the new product.
The next factual record will come from independent safety testing and age-assurance error data and school, parent and teen usage evidence after rollout. Until those records appear, the account remains bounded by the cited reporting, measurements and explicitly attributed statements.