What Claude KYC is: the three things at the core of the policy
KYC stands for Know Your Customer. Since April 2026, Anthropic has been rolling out identity verification on claude.ai: it is run by Persona Identities, a third-party identity provider, and asks you to capture a physical government-issued photo ID (passport, driver's licence or national ID card) and complete a live face check — a flow that takes a few minutes when it goes smoothly. Anthropic gives exactly three purposes: preventing abuse, enforcing its usage policies, and meeting legal obligations; the data collected is used only to confirm who you are.
It is not mandatory for everyone: you act only when a Verify your identity prompt appears in your account, and there is no official way to verify in advance. There are two paths — individual verification, and business verification on behalf of an organization. Industry context: Anthropic is the first major AI company to push identity checks at consumer scale, but not the first to do it at all — OpenAI verified API organizations back in April 2025, and in 2026 brought identity verification into ChatGPT itself through the Daybreak program. Verify first, then unlock stronger capabilities is becoming the industry's shared direction.
The policy timeline: from the April launch to the October tiers
Laid out in order, the public record shows the trigger surface expanding, not shrinking — learning the rules early beats scrambling when a prompt interrupts your work.
| When | What happened | What it means for you |
|---|---|---|
| 2025-04 | OpenAI introduced identity verification for API organizations | The industry precedent; consumers were not asked at the time |
| 2026-04-14 | Anthropic updated its Help Center and announced identity verification rolling out on claude.ai (run by Persona) | Certain features began requiring verification; privacy and regional-fairness debate followed |
| Mid-2026 | Users widely reported being asked to verify during Max / Pro subscription checkout (community and media reports) | Verification became tied to paid flows, widening the trigger surface |
| 2026-08-12 | The Help Center article Identity verification on Claude was updated | The current source of truth for documents, failure reasons and the appeal entry |
| 2026-10-06 | The Cyber Verification Program expanded into Defense / Red Team / Specialized tiers | Verification shifted from anti-abuse to unlocking cyber capabilities by trust level |
Who gets asked: four trigger scenarios and two paths
Triggering is decided by Anthropic's risk controls and integrity checks — it cannot be predicted or scheduled. Across the official article and community reports, triggers concentrate in four scenarios:
| Trigger | What happens | State until resolved |
|---|---|---|
| Accessing certain features or advanced capabilities | A verification prompt appears when you use the feature | The feature stays locked until you verify |
| Routine platform-integrity checks | Regular risk-control sampling that can reach any user | Follow the prompt to completion |
| Safety and compliance review | Triggered by suspected fraud or abuse signals | Some features are restricted until verified |
| Subscription and upgrade flows | Community reports concentrate on Max / Pro checkout and upgrades | The subscription flow cannot finish without verification |
Documents and privacy: who holds your ID and how
Anthropic accepts government-issued physical photo IDs from most countries: a passport, a driver's licence or state/provincial ID, or a national ID card — the exact list is whatever your current verification page shows. The document must be legible, undamaged and unexpired. Explicitly not accepted: photocopies, screenshots or scans, digital IDs in phone wallets (including mobile driver's licences), temporary paper documents, and student/employee cards or bank cards. The flow combines document capture with a live face check, so the device camera must work.
Privacy is the policy's biggest controversy, and here is how Anthropic handles it: your document and selfie are collected by Persona and stored on Persona's servers, encrypted in transit and at rest, never used to train models, and never shared with third parties for marketing; Anthropic does not copy or store the raw images, and only views verification records through Persona's platform in cases such as appeals. One caution of our own: handing an ID to any third party carries risk, so read Persona's privacy terms first — and only ever submit inside the official flow. Never send document photos to an individual claiming they can verify for you.
Fixing verification problems: the official failure reasons mapped, plus pass preparation
Anthropic lists five failure reasons: blurry photos, an unreadable document, an expired document, an incorrect registration number, and technical issues. Most are preventable before you capture anything. If the verification link (inquiry.withpersona.com) expired or will not open, go back to the official page, re-enter from the verification entry and finish in one pass with the fresh link — the symptom-by-symptom version is in inquiry.withpersona.com won't open, expired or failed? The fix guide. The checklist below is community experience, not an official promise — the official review always decides.
- Even lighting; no backlight or harsh top light; clean the lens first.
- All four corners of the document in frame, laminate glare avoided, text legible.
- Hat, sunglasses and mask off for the face check; face the camera.
- Grant camera permission in advance; close other apps that hold the camera.
- A stable network; do not switch IPs, refresh the page, or change devices mid-flow.
- Before starting, confirm the document is unexpired and of a type your page accepts.
| Official failure reason | Matching fix |
|---|---|
| Blurry photo | Even light, clean lens, steady hands — then retake |
| Document unreadable | All four corners in frame, no glare — then retake |
| Document expired | Use a document within its validity period |
| Incorrect registration number | Re-check the entry against the page's requirements |
| Technical issues | Switch browser or network and retry later; if the link died, re-enter the flow for a fresh one |
What failure means: a failed check is not a ban
Follow the page's prompt and retry — most capture problems pass on a clean second attempt. A failure is not a ban: you can usually still sign in, with some features gated. If the account itself is restricted or banned after verification, use Anthropic's official appeal form and the Safeguards team will review — that is not a scenario any product on this site handles.
If your documents are not accepted (for example the page does not take IDs from your issuing country), contact Anthropic through the help form linked in the official article — this store does not take such cases. The same applies if you cannot complete the face check in person: the official support channel is the only route.
The latest move: the Cyber Verification Program's three tiers (2026-10-06)
You cannot read where Claude KYC is heading without the October 6 announcement: Anthropic merged the original Cyber Verification Program and Project Glasswing into a single three-tier program. Verified security professionals get fewer cyber-related blocks on Claude Opus 5.5, Sonnet 5.5, Mythos 5.1 and future models. Applications go through the Verification Portal (sign in, then Programs → Cyber Verification Program), and Anthropic aims to send a first reply within seven business days — a decision or a request for more information. Member organizations must accept data retention so Anthropic can monitor for misuse, and Defense members must adopt phishing-resistant MFA by 2026-12-15.
| Tier | Who it is for | Review |
|---|---|---|
| Defense Access | Defensive work (SOC, incident response, malware reverse engineering, vulnerability analysis); security teams at companies, nonprofits, universities and government bodies, critical-infrastructure operators, smaller security firms, open-source maintainers, and individual researchers with a disclosure track record; individuals need a paid plan | Aims to reply within days |
| Red Team Access | Adds authorized penetration testing and red-teaming on top of Defense; organizations only (in-house red teams, government red teams, security and pen-test firms) | A few weeks; Defense is enabled in the meantime |
| Specialized Access | The fewest blocks; for verified organizations authorized to test safety-critical systems such as flight operating systems, power grids, telecom networks, interbank infrastructure and government networks; existing Glasswing members transition automatically | Reviewed in depth, currently in collaboration with the US government |
Frequently asked questions
Does every Claude user have to verify?
No. It is trigger-based: certain features, routine platform-integrity checks, safety and compliance reviews, and — per community reports — subscription checkout flows. Without a prompt there is nothing to do, and no official early-verification entry exists.
Can I verify Claude without a passport — with a national ID card?
Anthropic accepts government-issued physical photo IDs from most countries, and national ID cards are among the commonly accepted types — but the exact list is whatever your current verification page shows. If none of your documents are accepted, use the official help form; this store cannot take such cases.
Is it safe to hand my ID and selfie to Persona?
Per Anthropic's commitments: Persona hosts the data encrypted, it is not used for training or shared for marketing, and Anthropic does not store the raw images. Still, submitting an ID to any third party carries risk — read Persona's privacy terms and only submit inside the official flow.
What happens if verification fails? Is the account banned?
A failed check is not a ban — you can usually still sign in with some features gated, and the page lets you retry. If the account is restricted or banned after verification, use Anthropic's official appeal form; no product on this site handles that scenario.
Can it trigger again after I pass?
Yes — routine platform-integrity checks can re-trigger it, and that is up to Anthropic's risk controls. Keep a screenshot of the completed state; if it reappears, work through this checklist again.