Anthropic restricts Claude to users 18 and over with third-party age verification

Anthropic has turned its consumer Claude assistant into an adults-only product. According to a support article on the company's help centre, Claude is restricted to people over 18, new users must confirm their age during account setup, and existing users suspected of being underage now have to verify their age through a third-party service before they can keep using the product. The policy drew wide attention after the support page reached the front page of Hacker News on 11 September 2026.

How the restriction works

Anthropic says it operates safety systems designed to spot accounts that appear to belong to people under 18, and that it disables accounts based on activity patterns associated with minors. An affected user receives a notification email containing a verification link. Passing the check gets the account reinstated; the support article does not spell out what happens to accounts that fail the check or abandon the process.

Verification itself is handled by Yoti, an external age-verification platform. The documentation says Yoti offers several ways to confirm a user is 18 or older, so people can choose whichever method suits them, though the page does not enumerate the available options.

What happens to verification data

The data-handling arrangement is the detail most likely to concern privacy-minded users. Anthropic states that Yoti is independently audited for SOC2 compliance, and that selfies, document images and any personal data are deleted by Yoti as soon as the age check is completed. Anthropic itself receives only a pass or fail outcome: the company says it never sees the submitted ID or image, and does not process or store personal data from the verification.

Even so, the practical effect is that a user flagged as possibly under 18 must hand a selfie or identity document to an outside company to regain access to a chatbot. That is a materially different trust model from a self-declared age checkbox at signup.

Why it matters

The change formalizes an adult-only stance for one of the most widely used consumer AI assistants, and it does so with active enforcement rather than attestation alone. Two aspects stand out.

First, detection-based enforcement cuts both ways. Behavioral signals will presumably catch some genuine minors, but they will also produce false positives, and the only route back into a disabled account is a document or biometric check with a third party. Users who are unwilling or unable to complete that check effectively lose their accounts by default.

Second, it indicates where safety policy for consumer AI products is heading. Anthropic presents the restriction as a safeguard, and its help centre lists related guidance on minimum-age requirements, identity verification, and real-time safeguards for Claude Opus and Sonnet. Age assurance is already a familiar obligation for social platforms in a number of jurisdictions; extending it to AI assistants carries the same trade-off — clearer protection for minors and cleaner liability for providers, in exchange for friction and identity checks for the broader user base.

For under-18 users there is no child mode or parental-consent path in the published policy: Claude is simply off limits. The support article defines the limit in terms of the consumer product, and does not detail how enforcement differs across Claude's various apps and interfaces — an open question Anthropic has not addressed on that page.