OpenAI updates Model Spec with U18 teen protections, adds age prediction and family tools

On 18 December 2025, OpenAI said it has updated its Model Spec, the internal guide for model behaviour, with a new set of Under‑18 Principles to shape how ChatGPT engages with teen users. The company said the update aims to deliver safer, age‑appropriate responses for people aged 13 to 17, informed by developmental science and external expert input.

According to the company, the U18 Principles clarify how the existing Model Spec should be applied in teen contexts. OpenAI said the approach prioritises prevention, early intervention and clear communication, and has been previewed with organisations such as the American Psychological Association.

• Put teen safety first, even when it conflicts with other goals
• Promote real‑world support by encouraging offline relationships and trusted resources
• Treat teens like teens, neither condescending to them nor treating them as adults
• Be transparent by setting clear expectations

OpenAI added that ChatGPT should take extra care in higher‑risk areas, including self‑harm and suicide, romantic or sexualised roleplay, graphic or explicit content, dangerous activities and substances, body image and disordered eating, and requests to keep secrets about unsafe behaviour. The assistant is expected to offer safer alternatives and encourage offline help, and, where there is imminent risk, to point users to emergency services or crisis resources.

How will the protections change ChatGPT’s replies for teens

The company said teens should encounter stronger guardrails, clearer cautions and nudges toward trusted adults when conversations move into sensitive territory. OpenAI is also beginning to roll out an age‑prediction model on consumer plans so that teen safeguards can be applied automatically, defaulting to a teen experience when age is uncertain and giving adults ways to verify. Industry reporting the same day noted similar moves across the sector to detect and manage underage usage.

Parental controls and resources

OpenAI said the teen protections build on its multi‑layered safety work in ChatGPT. Parental controls, launched earlier this year, let parents link to a teen account and tailor settings. The company said it has extended protections to features such as group chats, the ChatGPT Atlas browser and the Sora app, and has added expert‑vetted resources for families, including a family guide and tips reviewed by ConnectSafely and members of its Expert Council on Well‑Being and AI. Break reminders during long sessions are designed to promote healthier use.

Expert input and partnerships

OpenAI said it worked with external experts while drafting the U18 Principles. The American Psychological Association reviewed an early draft and underscored the need for developmentally appropriate precautions and for balancing AI use with human relationships. OpenAI also referenced its Expert Council on Well‑Being and AI, its Global Physician Network, and a partnership with ThroughLine to surface local helplines in ChatGPT and Sora.

What comes next?

OpenAI said it will refine the U18 Principles over time as research, expert guidance and real‑world use evolve. The company plans to keep improving parental controls, expand resources for families and continue safety‑related model improvements.