OpenAI Updates Model Spec with U18 Teen Safety Principles for ChatGPT

OpenAI has up to date its Mannequin Specification—the rulebook governing ChatGPT’s conduct—with new “U18 Ideas” particularly designed to guard teenage customers. The December 2025 replace, developed with enter from the American Psychological Affiliation, establishes how the AI assistant ought to work together with customers ages 13 to 17.

The transfer follows OpenAI’s Teen Security Blueprint unveiled in November 2025 and comes as main tech firms face mounting strain over youth security. Meta expanded its personal AI security instruments for teenagers in October 2025, signaling an industry-wide shift towards age-differentiated AI experiences.

4 Core Commitments

The U18 Ideas relaxation on 4 pillars: prioritizing teen security even when it conflicts with different objectives, selling real-world assist and offline connections, treating youngsters appropriately moderately than as youngsters or adults, and sustaining transparency in regards to the AI’s limitations.

ChatGPT will now apply heightened warning when discussions with teen customers enterprise into high-risk territory. This contains self-harm, romantic or sexualized roleplay, specific content material, harmful actions, substance use, physique picture points, and requests for secrecy about unsafe conduct.

“APA encourages AI builders to offer developmentally applicable protections for teen customers of their merchandise,” mentioned Dr. Arthur C. Evans Jr., the group’s CEO. He emphasised that human interplay stays essential for adolescent growth and that AI use ought to be balanced with real-world connections.

Technical Implementation

OpenAI is rolling out an age-prediction mannequin throughout client ChatGPT plans. When the system identifies an account as belonging to a minor, teen protections activate robotically. Accounts with unsure or incomplete age information will default to the U18 expertise till grownup standing is verified.

Parental controls now lengthen to newer merchandise together with group chats, the ChatGPT Atlas browser, and Sora. The corporate has additionally partnered with ThroughLine to offer localized disaster hotlines inside ChatGPT and Sora, connecting customers to real-world assist when wanted.

Broader Trade Context

The replace displays rising regulatory scrutiny of AI interactions with minors. OpenAI’s Skilled Council on Properly-being and AI, established in October 2025, continues advising on wholesome AI use throughout age teams. A world community of physicians now helps consider mannequin conduct in delicate conversations.

For buyers watching OpenAI’s trajectory towards a possible public providing, the teenager security infrastructure represents each a compliance funding and a aggressive moat. Firms that set up strong youth safety frameworks early could face fewer regulatory obstacles as AI oversight tightens globally.

OpenAI acknowledged it would proceed refining these rules primarily based on new analysis, knowledgeable suggestions, and real-world utility information.

Picture supply: Shutterstock