Our approach to age prediction

Source: OpenAI Date: 2026-01-20 URL: https://openai.com/index/our-approach-to-age-prediction

Summary

Title-only: OpenAI describes their approach to predicting user age — likely covering AI-powered age estimation from behavior patterns, profile signals, or optionally provided identity information — to enforce age-appropriate content policies and comply with youth protection regulations. January 2026 places this alongside the EMEA youth wellbeing grant and the broader child safety push, suggesting age prediction is the technical infrastructure enabling their youth safety commitments.

Implications

The age verification thread. Age prediction without hard identity verification is technically difficult and legally fraught — models that estimate age from behavioral signals can be wrong, and the consequences of misclassifying a minor as an adult (or vice versa) are significant. OpenAI publishing their approach signals they’re investing seriously in age estimation as a platform capability, likely ahead of regulatory requirements that will mandate age verification for AI products accessible to minors.

Privacy vs. safety tradeoff. Behavioral age prediction creates a surveillance-adjacent capability: the same signals that estimate age also reveal usage patterns, preferences, and context. Publishing the approach is OpenAI being transparent about a capability that users might reasonably find concerning — the balance between protecting children and creating behavioral surveillance infrastructure requires ongoing public accountability.