Claude May Ask for Your ID to Restore Access

Anthropic updated its privacy policy on June 17, 2026, to allow Claude to require government-issued identity documents as part of an account appeals process. The change goes into effect on July 8, 2026. If your account is flagged for potentially fraudulent activity or a terms-of-service violation, Anthropic may now offer you an appeal path — one that involves uploading a passport or driver's license, submitting a selfie, and providing a facial geometry scan. A company spokesperson described the update as a change 'to the appeals process,' emphasizing that it applies to only a small subset of users who trigger account review.

The identity checks are handled by Persona, a San Francisco-based identity verification company that works with a range of consumer platforms. Persona determines how long it retains your documents and biometric data — Anthropic's updated policy does not specify deletion timelines. Because Persona operates as a US company, the information it holds is subject to requests from US government authorities. Anthropic describes the verification as a measure that allows compliance with state and country age-verification requirements, helps prevent fraud and abuse, and gives the company a mechanism to investigate policy violations where an account might otherwise simply be terminated.

Why this is bigger than it looks

An AI chatbot collecting biometric facial data and government-issued identification would have seemed like an extreme step two years ago. Today it reflects where regulatory pressure is actually headed. The EU AI Act, US state laws around online safety, and age-verification mandates in multiple countries are all pushing platforms toward knowing who they are talking to — not just a username and an email address, but a verifiable identity. The appeals-process framing keeps the scope narrow for now, but it is also infrastructure. An ID verification flow that exists for appeals can, without much additional engineering, be applied to age-gating sensitive features, meeting different jurisdictional requirements, or blocking coordinated misuse at a scale that anonymous accounts currently make difficult.

The timing adds another layer. Anthropic is currently in a visible dispute with the US government over access to its most powerful models, and is simultaneously under pressure from regulators in multiple directions. Building the ability to verify who users actually are — rather than relying on a self-reported account — gives Anthropic instruments it previously lacked: for compliance, for liability management, and for enforcing terms of service in situations where the consequences of violations are serious. None of that is spelled out in the policy update, but the capability now exists.