Claude Age Assurance Policy – Why Anthropic Is Blocking Minors

TL;DR – Claude is now an 18‑plus service

Anthropic has updated Claude’s consumer terms so that anyone under 18 is blocked from using the product unless they pass a Yoti‑powered age‑verification flow. The change is enforced automatically by internal safety signals and applies to all Claude accounts, including paid Pro/Max plans.

What the policy says

Age restriction – Claude is only available to users who are 18 years or older. The age check is performed during account creation and again if the system detects “minor activity.”
Detection mechanisms – Anthropic’s safety systems monitor language cues, usage patterns, and other signals that may indicate a user is a minor. When such signals appear, the account is temporarily disabled and the user receives an email prompting verification.
Verification provider – The verification step is handled by Yoti, a third‑party identity‑verification service. Users can verify via:
1. Facial age estimation – a selfie is analyzed to estimate age (no ID needed).
2. Document upload – a passport, driver’s licence, or national ID is photographed and uploaded.
3. Yoti Digital ID app – users who already have a Yoti account can share an “over‑18” attribute.
Data handling – Anthropic receives only a binary pass/fail result. Yoti deletes all personal data (selfie, document images) immediately after verification. Anthropic does not store any of the submitted PII.
Scope – The restriction applies to the consumer Claude product and, according to the support article, also to Team and Enterprise plans (though the article does not explicitly confirm this for enterprise accounts).

Community reaction on Hacker News

Main concerns

“The fact that Anthropic only receives a result, not the data itself, does not make me feel any better about this.” – mayhemducks

“Yoti is SOC2‑audited, but SOC2 is a self‑assessment framework that can be gamed.” – abustamam

“Why not let parents handle age verification instead of forcing users to upload IDs?” – scronkfinkle

“If the detection is based on language cues, it can be easily bypassed or produce false positives, as shown by a user who asked Claude to test the system.” – waltbosz

Points of agreement

Privacy worries – Several commenters highlighted that handing over government‑issued IDs to a private vendor raises data‑security concerns, especially after recent fines against Yoti for biometric‑data handling.
Effectiveness doubts – Users noted that age‑detection heuristics (spelling, phrasing, topics) are unreliable and can be gamed, making the policy feel more like a legal shield than a safety measure.
Industry inconsistency – Others pointed out that OpenAI, Google, and Meta allow users 13 + with parental consent, questioning why Anthropic adopts a stricter 18 +  rule.

Notable counter‑arguments

Legal liability – padolsey argued the move is primarily a liability protection for Anthropic rather than an ethical stance on model safety.
Educational impact – whazor and fnordpiglet expressed concern that blocking minors limits access to a powerful learning tool, especially for self‑taught students.
Technical workarounds – Some users suggested using non‑English or open‑source models to avoid verification, highlighting a potential shift toward decentralized AI.

Why Anthropic may have taken this step

Regulatory pressure – Recent legislation in the EU, Australia, and the U.S. (e.g., SB‑1119) is tightening requirements for AI services aimed at children. By enforcing an 18 +  cutoff, Anthropic sidesteps many of these obligations.
Risk management – Minors cannot enter binding contracts, which complicates enforcement of Terms of Service and liability for harmful outputs. An age barrier simplifies legal compliance.
Brand protection – Limiting access reduces the chance of high‑visibility incidents (e.g., minors prompting unsafe content) that could damage Anthropic’s reputation.

How the verification flow works in practice

Trigger – Either during sign‑up or after a detection event, the user receives an email with a verification link.
Yoti session – The link opens a Yoti UI where the user selects a verification method (selfie, document, or existing Digital ID).
Result – Yoti returns a simple "over‑18" flag to Anthropic. If the flag is positive, the account is re‑enabled; otherwise it remains disabled.
Data deletion – Yoti’s policy states that all uploaded media are purged immediately after the check. Anthropic stores no copy of the images or documents.

Potential implications for developers and enterprises

API usage – The policy applies only to the consumer UI. Claude’s API can still be integrated into products that serve users under 18, provided the developer assumes responsibility for age compliance.
Team/Enterprise accounts – While the support article does not explicitly clarify, many enterprise customers have reported that the same age‑check is enforced on shared workspaces, potentially affecting internal training tools.
Alternative solutions – Companies may migrate to open‑source LLMs or other providers that do not impose age restrictions, accelerating the adoption of self‑hosted models.

Bottom line

Anthropic’s new age‑assurance policy blocks anyone under 18 from using Claude unless they pass a Yoti verification. The move appears driven by legal risk mitigation rather than a purely safety‑first philosophy, and it has ignited a debate about privacy, efficacy, and the broader impact on education and competition in the AI space.