Anthropic Now Requires Photo ID + Selfie to Use Claude: What It Means and How to Prepare

Anthropic has quietly started asking some Claude users for a government photo ID plus a live selfie. Here is who is affected, why it is happening, and what to do if you get flagged.

On April 20, 2026, The Information reported that Anthropic has begun asking selected Claude users to verify their identity before continuing to use the service. The verification flow asks for two things: a photo of a government-issued ID and a live selfie taken on the same session so the two can be matched.

The rollout is not universal. Anthropic is targeting accounts whose signals match what the company calls “adversary risk” — traffic from the four US-designated adversary countries (China, Russia, North Korea, Iran), bulk account creation patterns, payment methods tied to embargoed jurisdictions, and API traffic that looks like model distillation. Everyone else continues to use Claude as usual.

Why This Is Happening Now

Two forces are pushing Anthropic in the same direction at the same time.

1. The US AI diffusion rule

Updates to US export controls in late 2025 put frontier AI labs on the hook for knowing who is using their most capable models. Closed-weight frontier models like Claude Mythos 5 now sit in roughly the same category as advanced semiconductors: the company that serves them is expected to use “know-your-customer” practices when the risk profile warrants it. Photo-ID + live-selfie is the least invasive verification flow that satisfies the rule.

2. Adversarial distillation campaigns

In its April 2026 threat report, Anthropic documented coordinated extraction campaigns against Claude by Chinese labs including DeepSeek, Moonshot, and MiniMax, using tens of thousands of fraudulent API accounts. The goal of these campaigns is adversarial distillation — cloning a frontier model's behavior into a cheaper open-weights replica, then selling that replica back into global markets at a fraction of the price. ID verification is the single most effective tool the company has to break that supply chain at the account-creation layer.

Who Gets Asked (and Who Does Not)

What the Verification Flow Looks Like

Users who are asked to verify see a single-page flow inside Claude.ai or the Anthropic Console. The steps are:

1. Select the country that issued your ID.
2. Upload a clear photo of a government-issued document — passport, national ID card, or driver's license.
3. Take a live selfie in-browser (webcam or phone camera).
4. Wait for the automated match to complete — typically under two minutes.
5. Continue using Claude with the same account.

Anthropic has stated that the ID and selfie are processed through a third-party KYC vendor, stored encrypted, and deleted after a defined retention window once verification succeeds. The raw images are not used to train any Anthropic model.

What to Do If You Are Flagged by Mistake

Every automated risk system produces false positives. If you believe you have been flagged in error — for example, you use a VPN for general privacy reasons, or you share an IP range with many other residential users — there are three clear options.

Option 1 — Complete the verification

This is the fastest path. If you are a legitimate user, uploading a real ID and taking a live selfie unlocks full access within minutes. Anthropic retains the verification result, not the documents themselves, so you will not be asked again unless new risk signals appear.

Option 2 — File an appeal

If you cannot or prefer not to verify, the flagged-account email includes an appeal link. Appeals are reviewed manually and typically resolve within 3–5 business days. Expect to provide billing history, a short description of how you use Claude, and any organizational context (company name, domain).

Option 3 — Keep working through Happycapy

Because Happycapy is an agent platform that reaches Claude through enterprise channels, end users never hit the ID-verification layer directly. You still get Claude-quality reasoning; you just avoid the KYC step entirely. Many users use Happycapy as their primary day-to-day interface and keep a direct Anthropic account only for edge cases like playground experimentation or raw-API work.

How This Compares to OpenAI and Google

Anthropic is the first major US frontier lab to roll out visible photo-ID-plus-selfie verification at the consumer account layer. The other two leaders are operating somewhere in between:

What This Signals for AI in 2026

Anthropic's ID rollout is an early, visible sign of something many have expected for two years: frontier AI is being re-classified from “consumer software” to “dual-use technology.” That does not mean most users will be inconvenienced — the overwhelming majority of people who ever use Claude will never see a verification prompt. But it does mean that the old “sign up with any email and go” pattern is ending at the frontier layer.

The practical takeaway for individuals and small teams is simple: keep more than one path to frontier AI. A direct Claude account for experimentation. A cloud- vendor tenant (Bedrock, Vertex, Foundry) for production. And an agent layer like Happycapy that stays friction-free and keeps your daily workflow moving when the frontier providers tighten a knob.

Further Reading

For more context on the events shaping this policy, see our coverage of the Frontier Model Forum's anti-distillation alliance, Claude Mythos 5's 10-trillion-parameter launch, and Anthropic's $30B Series G — the three stories that together explain why the company is tightening access now.

FAQ

Sources: The Information — “Anthropic Requires ID & Selfies to Block Adversary Access” (April 2026); Anthropic Threat Report, April 2026; Frontier Model Forum joint statement (April 7, 2026); Reuters — US AI diffusion rule updates (late 2025).