Character.AI named its age-check vendor without naming its accuracy

The design

On 29 October 2025, Character.AI published a post announcing that it would remove open-ended chat for users under 18 and, alongside that change, roll out new age-assurance functionality. The company's own account states that it built an age-assurance model in-house and paired it with third-party tools, naming Persona as the vendor it would combine with its own system. A follow-up post on 21 November 2025 states that the rollout began in the United States and would continue to other markets. Character.AI's help center describes the mechanics in more plain terms: Persona, described as a third-party identity-assurance company, handles the actual verification attempt, and an account is routed into either a full experience or a restricted under-18 experience depending on the result.

What the evidence says

The company's own announcements establish that a verification step exists and name the vendor, but neither post states how the model or Persona's checks perform, what error rate either produces, or how many accounts have been reclassified. The help-center article corroborates the partnership independently of the announcement, which strengthens confidence that Persona is genuinely involved rather than being an aspirational partner named once and dropped. None of the three documents is a regulator's or an auditor's assessment; each is the company's own description of its own system, so claims about how well it distinguishes minors from adults remain unverified by any outside party in the material reviewed here.

What it asks of people

A user asked to complete age assurance is asked to submit identifying material to a company that is not Character.AI itself, on Character.AI's representation that the process is necessary to receive an age-appropriate experience. The help center tells users that finishing the process does not guarantee a particular outcome, since placement depends on what the check returns rather than on completing the step. Parents are told, separately, that the rollout affects their teens directly by changing which features are available to them.

Privacy and safeguards

Character.AI's help center states plainly that Character.AI does not directly handle or store the documents submitted during verification, attributing custody of that material to Persona instead. That statement describes where responsibility is assigned, not what Persona itself retains or for how long, a gap the cited pages do not close. The announcement and update posts describe re-verification and support-desk correction for misclassified adults, but neither document discusses deletion timelines for the identity material once a check is complete.

• Has Character.AI published, or has Persona published, any accuracy or error-rate figures for this specific deployment?
• What does Persona's own retention schedule say about identity documents once a verification attempt concludes?
• How does an adult wrongly placed in the under-18 experience contest that result, and how long does correction take?

Taken together, the record shows a named vendor and a stated purpose, but not an independent measurement of how the system performs; that gap is exactly what a family or a researcher should keep separate from the company's own confidence in its design.