Building towards age prediction

Source: OpenAI Date: 2025-09-16 URL: https://openai.com/index/building-towards-age-prediction

Summary

Summary

OpenAI published research on building towards AI-based age prediction capabilities — exploring how AI systems can estimate the age of users or subjects in images or interactions to enable age-appropriate content filtering and compliance with minor protection requirements. The work addressed the technical challenge of reliably inferring age without requiring identity verification.

Implications

Safety/policy thread. Age prediction is a live regulatory requirement: the UK’s Age Appropriate Design Code, US state laws around minor protection online, and the broader child safety commitments OpenAI had made all require some mechanism to restrict minors’ access to adult content. Technical age estimation (from images, voice, or behavioral signals) is one approach, though it carries both false positive/negative risks and privacy concerns. The September 2025 timing follows OpenAI’s teen safety blueprint from November and other child safety initiatives, suggesting this is the technical research underlying those product commitments. The balance between privacy (not requiring ID verification) and accuracy (age estimates are probabilistic) defines the design space.