OpenAI Postpones Adult Chat Feature as Age-Prediction System Fails Accuracy Tests

OpenAI has announced a significant delay in rolling out its adult chat functionality for ChatGPT following critical concerns about the reliability of its age-verification technology. According to data from NS3.AI, the system’s age-prediction mechanism failed to properly identify users, incorrectly categorizing teenagers as adults in approximately 12% of cases—a substantial error rate that raises serious questions about user safety and content protection measures.

Critical Flaws in Age-Prediction Technology

The decision to hold back the adult chat feature represents a major pivot for OpenAI following formal warnings issued by the company’s Expert Council on Well-Being and AI in January. The council specifically flagged potential vulnerabilities that could disproportionately affect younger users and those from marginalized communities. With error rates reaching double digits, the age-prediction system clearly presents significant technical challenges that demand resolution before deployment.

Safety-First Strategy and Industry Implications

This postponement underscores OpenAI’s commitment to prioritizing user protection over rapid feature releases. The company’s decision reflects broader industry recognition that deploying content-gated features without robust safety mechanisms can create unintended consequences. By delaying the adult chat rollout, OpenAI is setting a precedent for how AI companies should approach ethically sensitive features—ensuring that age-verification systems meet stringent accuracy standards before affecting real users.

The postponement also highlights the technical complexity inherent in building trustworthy AI systems, particularly when attempting to make nuanced distinctions about user demographics based on limited information. As OpenAI continues refining its approach to adult chat, the industry will be watching to see whether improved technology can resolve these fundamental accuracy issues.