OpenAI has begun a phased launch of its ChatGPT user age verification system, in which some accounts are prompted to prove their age of majority with a passport or other government-issued ID, according to the company’s reference materials and user posts.
According to new OpenAI guidance, ChatGPT uses an age prediction model to assess whether a user’s account belongs to a user younger or older than 18. If the system believes the user is under 18, additional restrictions on sensitive content are automatically applied to the account, and for adults mistakenly classified as teenagers, there is an option to regain full access through document age verification.
In the published article Age prediction in ChatGPT OpenAI specifies that in case of erroneous attribution to the group under 18 years of age, an adult user can pass the verification by uploading a photo of a passport or other state ID card and selfies through the partner Persona: after confirming the age of the account is transferred to the standard “adult” mode. The company separately notes that the feature is “being rolled out gradually and may not yet be available” in certain countries.
A separate note Why am I being asked to verify my age? describes an additional scheme for jurisdictions where age verification is required by law: when logging into ChatGPT, such users are shown a banner asking them to verify their age within 60 days, otherwise access is blocked until the procedure is completed. For these verifications, OpenAI uses a third-party service called Yoti, which, depending on the method chosen, can request selfies, document uploads or verification via a mobile app.
The company explains that the development of the age prediction system and document checks is related to stricter requirements for the protection of teenagers when using AI services. In September, OpenAI announced the creation of a “separate experience” for users under 18, where ChatGPT must not engage in flirtatious conversations or discuss topics related to causing harm, and in cases of a clear threat to a teenager’s life, the system can notify parents or authorities.
OpenAI did not disclose the full list of countries and languages where the system is already active, limiting itself to wording about “gradual rollout.” At the same time, ChatGPT requests to provide IDs and selfies have been widely reported by users of English-language forums, where some of them have had their access to the full functionality of the service restricted until verification or proof of age is completed.