Anthropic Starts Checking IDs: Claude Now Asks for a Passport

Anthropic has quietly started requiring some Claude users to submit a government-issued photo ID and a live selfie before accessing certain features — a first for a major consumer AI chatbot. The company updated its support page this week confirming the rollout, which began appearing in user accounts over the last several days and was picked up by The Register on April 16.

Verification is handled by third-party identity vendor Persona, the same company Discord controversially used for age checks earlier this year. The rollout immediately sparked backlash across Hacker News and Reddit, with some users saying they'll cancel Pro and Max subscriptions rather than upload their passports to an AI product.

Context: From "Privacy-First" to "Show Me Your ID"

Anthropic has spent 2026 positioning Claude as the privacy-conscious alternative to ChatGPT. That pitch hinged on things like not training on user data by default, honoring deletion requests aggressively, and avoiding the kind of tracking that competitors lean on for ad-adjacent revenue.

Today's move cuts against that brand. Anthropic is now collecting the most sensitive identifier a tech company can hold — a passport image plus a biometric selfie — for use cases the company itself has not spelled out. The support page simply says users "might see a verification prompt when accessing certain capabilities," leaving open which features trigger the check.

Why this matters: regulatory pressure on AI is rising, and KYC-style checks are becoming one of the tools labs reach for to control dual-use capabilities like cyber and agentic computer use. But Anthropic is doing it first and doing it quietly, and that combination is what's drawing fire.

What the Policy Actually Says

According to Anthropic's updated support page, the verification process accepts a physical, undamaged passport, driver's license, or national ID card. Screenshots, scans, photocopies, photos of photos, digital/mobile IDs, and non-government IDs are all rejected. Some flows additionally require a live selfie captured through Persona's flow.

Anthropic lists several privacy guardrails: identity data won't be used to train Claude, Persona stores the images (not Anthropic), data is encrypted in transit and at rest, access is limited to account review and appeals, and Persona is contractually restricted to verification and fraud prevention only.

Anthropic's stated purpose is "to prevent abuse, enforce our usage policies, and comply with legal obligations." The company hasn't published a retention period or named the specific capabilities that trigger a check, though reporting suggests the rollout targets a small subset of accounts and features.

Persona, meanwhile, relies on a stack of subprocessors that includes AWS, Google, OpenAI, Stripe, and Twilio — a detail that's drawing particular attention from users who chose Claude specifically to keep their data away from some of those names.

Industry Impact

Anthropic is now the only major frontier AI chatbot asking for government ID at the consumer level. OpenAI's ChatGPT and Google's Gemini both run on account-based trust signals without requiring passport uploads. That gap creates a clear short-term competitive dynamic: users who bounce from Claude over the verification prompt land directly in competitors' funnels.

For the broader AI industry, the move likely previews the direction of travel. Age verification (for minors), high-risk capability gating (cyber, bio, agentic tool use), and regulatory compliance in jurisdictions like the EU and UK are all converging on some form of identity check. Anthropic is effectively running the first large-scale experiment on whether mainstream AI users will tolerate it.

Enterprise customers are mostly insulated. API-only integrations, Claude for Work, and Claude Enterprise handle identity through corporate SSO rather than Persona, so the passport flow is a consumer-product decision rather than a platform-wide one.

Expert and User Reactions

Reaction on Hacker News and Reddit has been sharply negative. Several threads pointed to Persona's February Discord incident — where age verification rollout sparked its own privacy backlash — as a reason to be skeptical of the vendor choice. Others flagged the timing: Anthropic shipped the change with no blog post, no email to users, and no pre-announcement on X, which many users felt was inconsistent with a company whose brand is built on transparency.

Help Net Security framed the decision as a deliberate trust test, arguing Anthropic is betting that users will accept verification if it's framed around abuse prevention rather than age gating. Decrypt was more pointed, noting the awkwardness of a company whose users switched from ChatGPT citing "surveillance fears" now asking for their passports.

There's also a practical concern: identity verification creates a new high-value target. A breach at Persona — or any of its subprocessors — would expose IDs belonging to Claude's most engaged users.

What's Next

The verification prompt is rolling out "case by case" rather than across the whole user base. Users who see it can either complete verification through Persona or use a different feature that doesn't require it. There's no published way to opt out of future prompts, though users can delete their account entirely.

Watch three things from here: whether Anthropic publishes a more specific list of which features trigger verification, whether it faces regulatory follow-up in the EU or California given the biometric data handling, and whether OpenAI or Google feels pressure to match rather than lean into the contrast.

Bottom Line

Anthropic just made the most privacy-visible policy change of any AI lab this year, and it did it without fanfare. If you use Claude and you've hit a verification prompt, you now know it's real, not a phishing test. If you're a Claude user on the fence about privacy, this is the week to read your account settings — and decide how much your passport is worth to keep talking to an AI.