ChatGPT Updates Model Rules to Strengthen Teen Safety Online

After a year when generative AI moved deeper into classrooms, homes, and everyday conversations, questions around how these systems engage with younger users have become harder to ignore. ChatGPT’s latest update signals a shift from broad safety assurances to more explicit, age-aware rules of engagement.

OpenAI has updated its Model Spec—the internal framework that defines how ChatGPT should behave by introducing dedicated Under-18 (U18) Principles. The changes are designed to guide how the AI supports teenagers aged 13 to 17, recognising that minors interact with technology differently from adults and require additional safeguards.

The update clarifies how existing model principles should be applied in teen-specific contexts, particularly in situations involving elevated risk, vulnerability, or long-term developmental impact.

Why Teen-Specific AI Rules Are Emerging Now

The introduction of U18 Principles reflects a broader reality: AI tools are no longer fringe utilities but part of daily digital life for teens. From homework help to creative expression, young users increasingly rely on conversational AI, often without the context or guardrails that traditional education or supervision provides.

According to OpenAI, the U18 Principles are grounded in developmental science and emphasise prevention, transparency, and early intervention. The company previewed the framework with external experts, including the American Psychological Association, as part of its effort to incorporate independent guidance into product design.

While the core Model Spec continues to apply to all users, this update makes clear that teen interactions require a more cautious and calibrated approach.

Four Commitments Guiding Teen Interactions

At the centre of the update are four guiding commitments that shape how ChatGPT should respond to teen users:

• Put teen safety first, even when it conflicts with other objectives

• Promote real-world support by encouraging trusted offline relationships

• Treat teens like teens neither condescending nor positioning them as adults

• Be transparent by setting clear expectations about limitations and intent

These commitments influence how the model navigates sensitive topics and how it balances helpfulness with restraint.

Stronger Guardrails for High-Risk Conversations

The revised Model Spec outlines areas where additional care is required when interacting with minors. These include self-harm and suicide, sexualised or romantic roleplay, explicit content, dangerous substances or activities, body image concerns, disordered eating, and requests involving secrecy around unsafe behaviour.

In higher-risk scenarios, the assistant is expected to offer safer alternatives and encourage teens to seek trusted offline support. Where there is imminent risk, the model is designed to guide users toward emergency services or crisis resources.

The update aligns with ChatGPT’s existing Teen Safety Blueprint and builds on safeguards already in place through parental controls and age-based content protections.

Parental Controls and Product-Level Changes

Alongside the Model Spec update, OpenAI has expanded teen safety protections across newer products, including group chats, the ChatGPT Atlas browser, and the Sora app. Parents can tailor how their teens use ChatGPT as new features are introduced.

To support family conversations around AI use, OpenAI has also added expert-reviewed resources to its parent resource hub. These include a Family Guide to Help Teens Use AI Responsibly and practical guidance for parents on discussing AI with their children.

Within the product itself, features such as break reminders are designed to promote more intentional and balanced use, particularly during extended sessions.

Expert Input Shapes the Framework

External expert engagement plays a central role in the updated approach. The American Psychological Association reviewed an early draft of the U18 Model Spec and emphasised the importance of developmentally appropriate protections.

“APA encourages AI developers to offer developmentally appropriate precautions for youth users of their products and to take a more protective approach for younger users… Youth experiences with AI should be thoroughly supervised and discussed with trusted adults,” said Dr Arthur C. Evans Jr, Chief Executive Officer, American Psychological Association.

OpenAI has also established an Expert Council on Well-Being and AI and continues to draw on clinical expertise through its Global Physician Network to improve how ChatGPT identifies distress and guides users toward professional care.

The company is beginning to roll out an age prediction model across ChatGPT consumer plans. When age signals are unclear, the system will default to a U18 experience, with options for adults to verify their age.

OpenAI says it will continue refining the U18 Principles as new research, expert feedback, and real-world usage patterns emerge, positioning teen safety as an ongoing responsibility rather than a one-time policy update.