OpenAI on Tuesday launched ChatGPT for Teens, an age-gated version of its chatbot that automatically routes users aged 13 to 17 into a restricted mode blocking self-harm, suicide, eating-disorder, violence, and sexually explicit content, along with romantic language and any suggestion the model has feelings or consciousness.

Enrollment isn’t opt-in. Users who self-report as minors, or whom OpenAI’s age-prediction system estimates to be under 18 based on query signals, land in the teen experience by default. OpenAI doesn’t verify ages, and likened its approach to Meta’s teen-account model on Instagram, an implicit concession that the industry has settled on inference rather than ID checks as the compliance floor.

The product design reads as a direct response to legal exposure. The launch follows lawsuits over teen safety, including one from the family of a 16-year-old who died by suicide after lengthy ChatGPT interactions. A 2025 Common Sense Media study cited by the Associated Press found that more than 70 percent of U.S. teens have used AI chatbots for companionship, the behavior the new safeguards are engineered to interrupt.

Allison Mishkin, head of child development at OpenAI, told the AP the team worked backward from that dynamic: “We went through and identified what are the hypothetical cues that a model could give that might make a teenager kind of develop a relationship to it.” Hence the ban on terms of endearment and consciousness claims, framed internally as reducing anthropomorphism and emotional dependence.

The rest of the feature set follows the same logic. Study Mode redirects homework shortcuts into guiding questions and step-by-step prompts, and “responsible homework reminders” flag apparent cheating. Parents who link accounts can set Quiet Hours and receive notifications in limited high-risk situations, including certain eating-disorder cases. OpenAI also announced a partnership with CodeAI on AI literacy for students and educators.

The launch arrives years after teens started using the product. That sequencing, product first, safety architecture later, is itself the story.

Sources