Start your 3-day free trial
Sign up to experience all premium features at no cost.
*Available only to new users. Each user is limited to one trial.


For Claude in China, the short answer is that it is not currently offered for commercial access in the mainland under Anthropic's published supported-regions policy. That answer applies before any network test: loading a page, changing an IP address, or possessing an account created elsewhere does not make an unsupported person or organization eligible. If you are an otherwise eligible traveler or employee diagnosing an existing, authorized account, separate Claude.ai, the API, organization policy, billing, account status, and the network path before acting.[1][2]
Key Takeaways:
- Check Anthropic's current supported-region and ownership rules before troubleshooting connectivity.
- Treat Claude.ai, the API, a third-party integration, and an organization workspace as different surfaces.
- Keep country, organization, identity, phone, and billing information accurate; a route change cannot alter eligibility.
- Stop if the proposed test would bypass a provider restriction, employer control, or applicable law.
This guide does not show how to evade Anthropic's regional policy. It provides a safe evidence sequence for distinguishing an official eligibility decision from an account, product, or network symptom.
“Available” can describe several different outcomes. A public marketing page may load while sign-in is unavailable. An existing account may authenticate while a particular model, paid plan, API key, or organization is unavailable. A third-party product may mention Claude while using its own service agreement and account boundary. Record the exact surface instead of converting one result into a claim about the whole platform.
| Layer | Question to answer | Evidence that belongs here |
|---|---|---|
| Official region | Does Anthropic list the user's and organization's jurisdiction? | Current supported-regions page and terms |
| Account | Is sign-in, verification, or recovery complete? | Exact account message and official support response |
| Product | Is the issue Claude.ai, API, plan, model, or workspace? | Product-specific screen, documentation, and status |
| Organization | Is access permitted by ownership and administrator policy? | Organization settings and authorized administrator |
| Network | Does the same authorized request differ by trusted route? | Time-stamped comparison with all other variables fixed |
Anthropic publishes separate lists for Claude.ai and commercial API access and reserves additional rules for organizations connected to unsupported regions. Its 2025 policy notice says the restriction also covers certain entities controlled from unsupported jurisdictions. Read both the current list and the applicable terms; do not infer eligibility from incorporation, travel, a reseller, or a successful page load.[1][2]
Open Anthropic's official supported-countries page and locate the relevant jurisdiction under the exact product surface. Check where the individual ordinarily resides, where the organization is based or controlled, and which legal entity would buy the service. If mainland China remains absent or the ownership rule applies, stop. A network change is not an official exception.
Save the page title, URL, check date, and the section you used. Supported lists can change, so a screenshot or memory from a previous trip is not enough. If your employer supplies Claude through an approved enterprise arrangement, ask its authorized administrator which product and entity are covered rather than assuming a personal Claude.ai rule applies.
Write down whether you are opening claude.ai, using an official mobile app, calling Anthropic's API, entering an organization workspace, or using a separate application that embeds a model. These paths can use different accounts, regions, billing arrangements, credentials, and administrators. Do not enter Anthropic credentials into an unofficial page simply because it uses the Claude name.
For API work, distinguish DNS or TLS failure from an HTTP response, authentication failure, permission error, credit or billing problem, model entitlement, rate limit, and safety response. Never paste an API key into a screenshot or diagnostic site. Rotate a key through the official console if it has been exposed.
Use the account's real country, identity, age, phone, payment, and organization information. Check whether an email verification, recovery step, administrator invitation, billing action, or terms acceptance is pending. If an organization owns the account, ask its administrator whether the member, domain, product, and model are enabled.[3]
Do not create replacement accounts, borrow another person's login, invent an address, or move billing details to make an error disappear. Repeated sign-in attempts from changing locations can add fraud or security signals and make the original diagnosis harder. Preserve the full error message and use Anthropic's official support path when the decision is account-specific.
Check Anthropic's official status page for the affected component and time window. A service incident can affect login, conversations, uploads, billing, or API operations differently. If the status page reports an incident, wait and retest the same operation; changing account or network settings adds noise.
If no incident is reported and the account is eligible, compare one public page and one authorized action on the current network. Record DNS result, whether TLS completes normally, the HTTP or app error, device, client version, and time. Never disable certificate validation, install an unknown root certificate, or use a copied login page.
Use the same device, account, Claude surface, request, and short time window. Compare the current network with one other trusted network. If applicable law, Anthropic's terms, and your organization permit it, a VPN can provide a controlled alternative network path; it cannot create regional eligibility, organization approval, billing entitlement, or account access. For an eligible account whose comparison is permitted, AethoVPN can supply that encrypted route; Anthropic's terms and local law still apply.
A changed result shows only that the route may influence that specific request. It does not prove that Claude is generally supported in China, that every feature works, or that continued use is permitted. Return to the baseline, record the result, and remove temporary diagnostic settings.
| Observation | Most useful next check | Do not conclude |
|---|---|---|
| Region or unsupported-location message | Supported-region policy and applicable terms | That rotating IP addresses creates eligibility |
| Sign-in or verification failure | Recovery state, identity request, official support | That the network is the only cause |
| Workspace opens but a model is missing | Administrator, plan, model entitlement, rollout | That Claude is unavailable everywhere |
| API returns 401 or 403 | Key validity, organization, role, policy | That DNS or routing failed |
| API times out before an HTTP response | DNS, TLS, proxy, firewall, service status | That the account is accepted |
| Billing fails | Supported billing entity and official billing support | That another route or card should be tried repeatedly |
Keep observed facts and interpretations separate. “The public page loaded at 10:15 on one connection” is defensible; “Claude works in China” is not. Likewise, a timeout on one hotel network cannot establish an official country restriction without the provider's policy.
Send official support or your organization administrator the product surface, account type, organization name without secrets, approximate location, time and time zone, device and client version, exact error, and whether the same authorized action differed on a second trusted network. Redact API keys, recovery codes, payment details, private prompts, uploaded documents, and customer information.
Ask a narrow question: whether the account and organization are eligible for that surface, whether a known incident affects it, or which verification step is incomplete. Do not ask support to validate a workaround that contradicts the published region policy.
Not under Anthropic's current published supported-regions list. Recheck the official Claude.ai and API sections because policies can change, and consider any organization-ownership restrictions that apply.[1][2]
No. A VPN changes the network path and visible exit IP. It does not change residence, organization control, identity, billing, terms, or the provider's eligibility decision.
Public pages, sign-in, account eligibility, subscriptions, workspaces, models, and API calls are different surfaces. A loaded public page proves only that the page was reachable at that moment.
Anthropic presents them as separate sections, so check the exact product you intend to use. API credentials, organization membership, billing, and model permissions add further checks.
Confirm the account and organization remain eligible, check the official status page, then record the exact surface and error. Compare one other trusted network only if terms, policy, and law permit it.
No. Keep country, identity, organization, phone, and payment information accurate. Use official support if the account contains outdated legitimate information.
Send the surface, account type, time, device, client version, exact error, and a redacted network comparison. Never include passwords, API keys, recovery codes, payment data, or private prompts.
Disclaimer: This article provides general access-diagnostic information, not legal advice or permission to use a service. Follow Anthropic's current terms, your organization's policies, and applicable law.
Sources checked 12 September 2026.
Related reading:
Sign up to experience all premium features at no cost.
*Available only to new users. Each user is limited to one trial.