Anthropic has published the operating details of an age policy that makes Claude an adults-only product: users must confirm they are 18 or over at sign-up, automated classifiers hunt for signs that an account belongs to a minor, and accounts showing those signs are disabled. The support documentation, which surfaced on Hacker News this week, also spells out the appeal route β and that route exists because the detector has been catching adults.
Key takeaways
- Claude's consumer product is restricted to people over 18, enforced by classifiers that scan for indicators of minor activity and disable accounts.
- Suspended users can restore access through Yoti using a facial age estimate, an ID photo, or a stored digital identity attribute.
- Anthropic says it receives only a pass or fail result and never sees the selfie or document.
How enforcement actually works
The age bar itself is not new. Anthropic announced in December 2025 that every Claude user must be 18 or older and must affirm that at registration. What the newly published support article adds is the enforcement layer behind that affirmation.
The company says it runs safety systems that look for signals an account is being used by someone under 18, and that it disables accounts on the strength of those indicators. Its initial classifiers flagged conversations in which users explicitly described themselves as minors. Anthropic has said it is extending that to subtler signals, which is where the accuracy question starts.
Suspension arrives by email. Affected users have reported wording telling them the company found signals the account was used by a child, that this breaks the rules, and that access has been paused. A verification link is attached, and it expires after 30 days.
The Yoti appeal, and what it costs you
Reinstatement runs through Yoti, a third-party age-assurance vendor, and there are three paths. A facial age estimate takes a selfie and infers whether the subject is over 18, with no document required. Alternatively users can photograph a passport, driving licence or national ID card, or share a pre-verified over-18 attribute from Yoti's digital ID app.
The privacy design is deliberately narrow. Yoti is independently audited to SOC 2, deletes the selfie, document images and associated personal data once the check completes, and passes Anthropic nothing but a binary result. Anthropic states it never sees the ID or the image and does not store personal data from the process.
That is a genuinely restrained architecture for identity checking. The awkwardness sits earlier in the chain: the trigger for all of this is an automated read of what users have been typing into a generative AI assistant. Users wrongly flagged have said as much, with one Pro subscriber posting that the company had seen their conversations and locked them out. Others reported regaining access after submitting ID, and at least one said their subscription was refunded while their saved projects were left broken.
Two strategies for the same problem
Anthropic's approach is a clean exclusion: no minors on the consumer product, with a separate set of guidelines governing organisations that serve minors through its models. That diverges sharply from the alternative now in the market, which is to build a distinct experience rather than a wall β the path in OpenAI's separate ChatGPT for teens, which bets on age assurance over ID checks.
Both are responses to tightening regulation, and platform-level rules are closing in from a third direction: app store age checks in some US states may independently block under-18 users from creating or signing into Claude on mobile. The unresolved question for Anthropic is the one put to it publicly and not yet answered β what the false positive rate of its classifiers is, how often suspensions are overturned, and how much of a user's conversation history a human reviewer reads before pulling the switch.
FAQ
What happens if Claude wrongly flags my account?
Access is paused and a verification link arrives by email, valid for 30 days. Completing the Yoti check and passing restores the account. Users who have gone through it report reinstatement after verifying, though some have described lingering damage to saved projects.
Does Anthropic see my ID or selfie?
No, according to its documentation. Yoti performs the check and transmits only a pass or fail result. Yoti deletes the images and personal data once the age determination is made, and Anthropic says it does not process or store any of it.
Can anyone under 18 use Claude legitimately?
Not the consumer product. Anthropic publishes separate guidance for organisations that serve minors using its models, so supervised access built on top of the API is treated differently from an individual signing up directly.
