November 2026 Claude bans and KYC checks: why KYC is only an entry ticket, how IP, DNS/WebRTC leaks, shared nodes and device fingerprints trigger bans, and what to check before paying.
The short version: KYC is an entry ticket, not a guarantee against bans. Most Claude suspensions come from the everyday egress environment, device fingerprint and behaviour patterns. Passing KYC, buying a residential IP or warming up an account does not fix those gaps.
Since November 2026, KYC checks and ban reports around Claude have surged. Some users paid hundreds of yuan for someone else to pass KYC; others bought expensive "dual-ISP residential IPs". Either way, the account often did not survive three days. The question is not whether verification happened, but whether every login after it stays internally consistent.
This guide merges the most common 18 questions from community reports and reorganises them by cause, environment, device, payment and recovery, cross-checked against Anthropic's official documentation. Community samples show direction, not official probabilities, and are not a stability guarantee.
Claude's risk controls do not rely on a single signal. Account, network, device and behaviour signals are read together, and any clearly abnormal layer can trigger verification or suspension.
| Cause | Typical symptoms | Check first |
|---|---|---|
| Dirty network egress | Instant suspension on login; new accounts on the same IP die in sequence |
Bloccato in un passaggio? Scrivi all’assistenza. Gli ordini pagati si gestiscono anche dalla pagina ordine.
| Shared IP, ban history, DNS/WebRTC leaks |
| Identity does not match the account | Portal check on upgrade or renewal; suspension soon after verification | Whether the ID is yours, registration details, payer consistency |
| Device and behaviour linkage | Multiple accounts on one device, bulk registration, abnormal automation | Browser fingerprint, device history, shared use |
| Payment and region risk | Card binding fails; payment succeeds but the subscription never attaches | Issuer region, payment channel, billing address |
Think of it as three questions: Are you a real user in a supported region behaving normally? Is your network identity stable over time? Does your behaviour look like a farm of accounts? If any answer is doubtful, KYC only postpones the question instead of resolving it.
Claude's identity checks run through Persona. Personal verification only accepts primary government-issued photo ID: a passport, a driver's licence or state/provincial ID, or a national ID card, plus a live selfie holding the document. Digital IDs on a phone, student cards, employee badges, library cards and bank cards are not accepted.
A US visa is an entry permit, not a standalone identity document. If you select the United States and upload a visa, the check usually fails at document-type recognition before the face step. See Identity verification on Claude for the official description.
Community reports split both ways. Most recent cases were suspended shortly after submission, but there are counterexamples that passed and survived for more than a month. There is no official pass rate by country, and any "guaranteed pass" claim cannot be verified.
The more practical constraint is regional policy. Anthropic's supported-regions list does not include mainland China or Hong Kong, while Taiwan is included. Even a user willing to verify legitimately may hit regional limits at the entry point or later. For the detailed KYC picture, see our separate Claude KYC guide; this article does not repeat it.
Yes, and the risk is significant. Paid verification only gets the account through the door. Persona records verification details such as the document, video, originating IP and duration. If one identity verifies multiple accounts, those accounts can later be linked. Once you log in from a restricted region or a fingerprint that does not match, risk controls can trigger again.
Claims such as "KYC account, never banned" or "permanent verification" are marketing language, not verifiable promises.
Separate two questions: can you verify legitimately, and is this account still worth saving.
If you have your own document, a clean everyday environment and no multi-account linkage, following the official flow is reasonable. Failed verification can usually be retried, and a failed check by itself is not a ban.
If the account already shows abnormal logins, shared use, repeated region jumps or a dirty environment, do not feed a clean identity into an already contaminated account. That binds your identity to a risky account. If the check appeared right after a charge, review the refund rules for that payment channel before deciding.
A country mismatch alone is not a problem. Cross-border travel and living abroad are normal, as long as the egress IP is in a supported region and is dedicated and clean.
The real risk is inconsistency: jumping between supported and unsupported regions, or having the verification IP, payment IP and everyday login IP in three completely different places. Risk controls look for consistency, not the cover of your passport.
This is the core of avoiding Claude bans. Environment problems do not disappear because KYC passed once.
Start with what the words mean. Dedicated means only you use that egress. Residential means the IP is registered like a home connection. Risk controls care more about the first.
Many expensive "dual-ISP residential IPs" are subleased second-hand addresses with unknown history: labelled residential, possibly already abused. Conversely, a dedicated, clean datacenter IP with no ban history and a complete set of leak-prevention rules can work long term for some users. Self-hosted VPS providers such as BandwagonHost are often cited in community discussions, but that is experience, not an official recommendation.
The order should be: dedicated, then clean, then region-compliant, then the residential label.
Three checks, in order of reliability:
| Check | How | Reference |
|---|---|---|
| ASN ownership | Look up the IP on ipinfo.io or bgp.he.net | Usage Type "isp" (home/business broadband) is closer to a real user than "hosting" |
| Fraud score and tags | Scamalytics, ping0.cc | Lower Fraud Score is better; the community often uses under 15 as a reference. Treat Datacenter / VPN / Proxy tags with caution |
| Interaction signals | Search and log in to common services from that IP | Frequent CAPTCHAs usually mean the range has been abused |
These are community heuristics, not an Anthropic score. They help rule out obviously bad IPs but cannot guarantee you will pass risk controls.
Four common leak paths:
The check is straightforward. With the proxy on, open an IP detection page and confirm the egress IP, DNS resolver location and WebRTC address agree. Then turn the proxy off and compare, confirming there is no path that still reaches Claude without it.
In community reports, shared proxy services are a hotspot. If one user on an egress IP triggers risk controls, the whole IP can lose weight, and other users may be asked to verify or get suspended too. This is not a published mechanism, but the risk of many people sharing one egress is real.
Do not run a primary or paid account on a shared node. One person, one IP, one device and one environment is the cheapest isolation.
Community experience says the risk is high. IP ranges from Tencent Cloud, Alibaba Cloud and similar providers are heavily used to reach AI services, so they are easily tagged as datacenter or abusive. That does not mean an automatic ban, but using one as a long-term primary egress puts the account in a heavily contested, heavily labelled IP pool.
If you already have a clean dedicated IP, use it. If you can only use a cloud server, at least check the ASN, fraud score and history, and configure complete leak-prevention rules.
Keep the system time zone, browser language and node location consistent. A common approach is Asia/Taipei or the node's region, with the browser language matching how you normally use the account.
Do not change time zones frequently to "look local". Frequent changes are themselves an anomaly signal. Stability matters more than disguise.
For one account on one device, no. The value of an anti-detect browser is isolating cookies, cache and browser characteristics across accounts, not becoming invisible.
If the IP is dirty, an anti-detect browser will not save you. If several accounts share one browser profile, even a good IP gets linked. The order remains: clean IP, then a separate environment, then fingerprint isolation.
Avoid switching several Claude accounts on your main device, especially after one of them has been banned. Cookies, site data and browser fingerprints can all become linkage evidence.
After a ban, change the IP and clean or rebuild the environment. Deleting cache without changing the IP and registering again often triggers another suspension.
Community reports describe OAuth credentials from a subscription being connected to third-party clients and used to run automation on subscription quota, followed by account restrictions. Implementations differ, so risk is not uniform, but two points are clear:
If you use Claude Code or another third-party client, confirm it falls within permitted use before keeping your main account connected long term.
Almost no benefit. The main triggers are the payment action itself plus the environment and account weight at that moment, not the age of the account.
In a dirty environment, a month of warming still ends in a check at payment or within two weeks after. In a clean environment, new accounts buying Pro directly also pass normally. Time is better spent on an environment audit.
No. The key is not a "US card" but whether the issuer is in an Anthropic-restricted region. Cards from mainland China, Hong Kong and other restricted regions are not advisable. Physical or compliant virtual cards issued in supported regions such as the US, UK, Japan, Singapore, Europe or Taiwan have worked.
Bybit is often mentioned because it supports USDT deposits and issues from a relatively compliant region, but other cards from comparable regions also work. Avoid BINs that have been heavily abused and flagged by risk controls.
The same applies to phone numbers. An overseas physical SIM or eSIM that can receive SMS long term is more stable. Virtual ranges from SMS-receiving platforms are easily blacklisted and may not receive secondary verification.
Submit an appeal through Anthropic's official support channel. Include the account email, subscription status, when it happened, why you believe it is a mistake, and order or payment records. Do not open multiple tickets in a row, and do not keep changing IPs to log in during the appeal.
Distinguish three states: failed verification, which can usually be retried; pending verification, where the account is asked for more identity information; and suspended or banned, where the account is disabled. The paths differ, and an appeal is not guaranteed to restore access.
It depends on the payment channel:
In community reports, instant refunds after a ban are less common than before. Confirm before buying: whether bans are covered, whether KYC status changes are warranted, and how iOS payments that never attach are handled. Sellers who promise "never banned" rarely deliver.
Before logging into your main account, confirm each item:
Dirty egress, shared IPs and identity-account mismatches are the three most common groups. KYC addresses identity only; it does not fix IP or device linkage.
Not necessarily. Email, IP, device fingerprint and registration behaviour can all trigger it. Changing only the email usually leads to another ban.
Risk is not split simply by plan. In community reports, Pro-to-Max upgrades trigger KYC more often, and Max verification paths are more complex, but bans still depend on environment and account history.
Usually not. A failed check can be retried and is not itself a ban. Violating the Usage Policy is a separate path.
No. iOS and Google Play only change the payment risk path. Community reports include store subscriptions that triggered verification, and iOS currently also has cases where payment succeeds but the subscription never attaches.
Hong Kong is not on Anthropic's supported-regions list, so using it as a long-term egress is high risk. Taiwan is on the list, but still requires a dedicated, clean IP.
Not by changing the device alone. An IP with a ban history may already be blacklisted; reusing it raises the chance that new accounts get linked. Change the IP as well.
No ready-made account is "safe"; only the risk boundaries are clear or unclear. Before buying, confirm whose document was used, whether multiple accounts are linked, whether bans are covered, and whether KYC status changes fall under after-sales.
Sometimes an email arrives; sometimes login simply shows suspended. Check the account mailbox and spam folder regularly, and keep payment records for appeals or refunds.
In one sentence: keep network, device, payment and identity consistent over time. A dedicated clean IP, a stable device environment, a compliant payment method and your own genuine verification beat any "anti-ban trick".
The structure of this article references a community summary by X user @ryuu_restart and was rewritten and cross-checked against the following official sources:
Details about IP scoring thresholds, shared nodes, cloud-provider nodes and third-party clients come from community reports and are experience-based, not official rules. Anthropic's supported regions, verification scope and risk controls may change. This article is not a stability guarantee and does not provide ways to bypass KYC, forge identity or evade regional restrictions. Using someone else's document or paying for verification may violate the terms of service and create account and financial risk.