Updating our Model Spec with teen protections

Source: OpenAI Date: 2025-12-18 URL: https://openai.com/index/updating-model-spec-with-teen-protections

Summary

OpenAI model spec update from December 2025 formalizing specific protections for teen users — users under 18 on the consumer ChatGPT platform. The protections likely included stronger defaults around age-appropriate content, restrictions on certain conversation types with minors, and parental notification or control mechanisms. This followed a period of significant public and regulatory pressure around AI companies’ obligations to child safety, including Meta’s legal exposure from teen mental health research and concurrent congressional hearings.

Implications

Regulatory preemption. Publishing child safety model spec updates in December 2025 positions OpenAI ahead of likely legislative requirements in the US and EU. The Children’s Online Privacy Protection Act updates (COPPA 2.0 discussion) and the EU’s DSA enforcement against major platforms created a political environment where proactive teen safety measures were necessary for OpenAI to maintain its consumer product.

Teen AI literacy resources companion. The teen protections update was published alongside the “AI literacy resources for teens and parents” post (December 2025), suggesting a coordinated communications push — not just restrictions, but educational framing. That combination is more defensible to parents and policymakers than restrictions alone.

Thread: consumer AI safety governance. Sits alongside the SafetyKit deployment (September 2025), the mental health grants (December 2025), and the sensitive conversation strengthening update (October 2025) as OpenAI’s consumer safety posture through late 2025.

Watch: Whether the teen protections translate into measurable behavior changes in the model when interacting with under-18 accounts, or whether they’re primarily policy documentation without implementation depth.