OpenAI rolls out age-prediction feature on ChatGPT

OpenAI announced the rollout of an age prediction feature on ChatGPT consumer plans to estimate whether an account belongs to someone under 18, aiming to apply appropriate safety measures.

The system uses behavioral signals, account age, activity times, and stated age to assess user age, with ongoing refinements based on data.

Accounts estimated to be under 18 will automatically receive restrictions on sensitive content, including violent, sexual, or harmful material, to protect minors.

Users incorrectly placed in the under-18 group can verify their age via a secure identity check through a selfie.

Parents can also customize teen accounts with controls such as quiet hours and distress notifications.

The feature is being introduced gradually, starting in the EU in the coming weeks, with ongoing improvements planned.

OpenAI said it will continue working with experts to enhance teen safety measures and share updates on their progress.

🔗 Source: OpenAI

🧠 Food for thought

Implications, context, and why it matters.

Crucial details on model accuracy and data privacy need to be clarified

The models empirical accuracy and possible demographic bias should be tested. Behavioral and account-level signals are used, such as account age, typical active hours, usage patterns over time, plus a users stated age, to predict whether an account likely belongs to someone under 18. No performance metrics are provided, yet they are needed to judge effectiveness under rules like the EUs Digital Services Act (DSA) 1.
Details of the selfie check run through Persona, an identity-verification provider, should be verified. It is pitched as a quick way for users to confirm age and regain full access. Data collection and retention rules are not spelled out, even though they are needed to assess privacy risk and data minimization under the UKs Childrens code (a privacy code of practice for online services likely to be accessed by children) 2.
Alignment with global standards that define highly effective age assurance has not been established. A high bar is set for some services under the UKs Online Safety Act. Direct biometric collection is discouraged by French regulators when alternatives are available 3.

A market emerges for specialized age-assurance tools in AI

AI service builders can treat OpenAIs in-house approach as proof of an opening for third-party age-assurance kits built for other AI products. Vendors must handle uneven rules across regions, including the UKs Online Safety Act and US state mandates 3.
Product teams can build privacy-preserving checks that avoid collecting identity data. French regulators prefer options like cryptographic proofs of age (methods that can confirm someone is above or below an age threshold without revealing their identity), which creates room for specialists 3.
Investors can back firms that track local deadlines, including the EUs DSA guidelines and the UKs phased rollout of its Online Safety Act 1 4. Providers selling compliance-as-a-service (outsourced tools and processes to help companies meet regulatory obligations) for AI can grow as enforcement ramps up.