AiStaffo

Anthropic's new regions policy bars China-controlled firms from Claude

Stock photo: laptop desk office notebook
Photo: Sora Shimazaki / Pexels

Anthropic's Supported Regions Policy took effect on 12 November 2026. It bars use of Claude by people in mainland China, by firms registered or headquartered there, and by organisations directly or indirectly controlled from unsupported regions, wherever their staff sit. The rules cover Claude.ai, Claude Code, Cowork, the API and Claude models sold through cloud platforms and authorised resellers. Businesses that run automated workflows on Claude should check where users log in, where the account holder is registered, and who owns the company. Accounts showing signs of use from an unsupported region may face identity checks, and accounts that fail them are banned, according to reports on the policy.

In short

  • Anthropic's Supported Regions Policy took effect on 12 November 2026.
  • It covers people in mainland China, firms registered or run from there, and organisations controlled from unsupported regions.
  • Claude reached through cloud platforms and resellers is covered too.
  • Check user locations, company registration and ownership before the next workflow review.
  • Accounts that fail identity checks are reportedly banned, so keep a fallback for each critical workflow.

What happened

Anthropic's Supported Regions Policy took effect on 12 November 2026. Under the policy, the company's products are available only in the countries and regions on its published list. Mainland China, Hong Kong, Macau and Russia are not on that list, according to the company's supported-countries page and to news reports on the change.

The policy covers Anthropic's apps, including Claude.ai, Claude Code and Cowork, its API and developer platform, and Claude models offered through cloud platforms and authorised resellers. Technology.org reports that the list names about 185 countries and regions, and that any place not on it counts as unsupported.

The change builds on a rule Anthropic introduced in September 2025. That earlier rule barred companies more than 50% owned from unsupported regions. Tom's Hardware, reporting on the 2025 decision, said it covered all Claude models and developer-facing tools.

Some Russian outlets reported a new ban on Russian employees from 12 November. Meduza compared the list dated May 2026 with the November 2026 version and found them identical, concluding that the country list itself had not changed. The new text mainly spells out how the existing rules are enforced.

Why it matters for businesses that automate work

Many automated workflows call Claude through an API, a cloud marketplace or a software vendor that resells it. Each of those routes falls under the policy, not only the company's own login. A workflow that runs well from an office in a supported country can stop working if the company behind it is controlled from an unsupported region.

Staff travel matters as well. The policy covers people who are physically in an unsupported region, even when their employer is based somewhere else. A finance manager who logs into a Claude-powered tool from a hotel in Hong Kong is within the policy's scope.

What changes in practice

The rules turn on three tests. The table below sets them out with a simple example for each.

TestWhat it looks atExample
Physical location of usersWhere the person is when they use ClaudeAn employee working from Macau on a laptop
Registration or headquartersWhere the company is registered or run fromA subsidiary registered in Russia, with staff in Canada
Ownership or controlWhether an unsupported region holds a majority stake, directly or indirectlyA firm whose parent company is majority-owned from mainland China

For most businesses, the practical work is a short review. The steps below follow the order a consultant would usually take.

  1. List every workflow that calls Claude, whether directly, through a cloud platform, or through a software vendor.
  2. Record the countries where the account holders and day-to-day users sit, including staff who travel.
  3. Confirm who owns the company that holds the account, and trace the ownership chain as far as the records allow.
  4. Ask each software vendor which AI model it uses and whether that model is reached through Claude.
  5. Name a fallback for each critical workflow, including a manual process that a person can run for a few days.

According to reports on the policy, accounts that show signs of use from an unsupported region may face identity checks, and accounts that fail them are banned. A banned account stops every automated job that depends on it at once. Business owners should test what happens to their reports, invoices or customer messages when a key account is suspended.

What automation cannot do here

Automation cannot settle ownership questions. It can collect the data, flag a workflow that depends on Claude, and show who uses it. Whether a company is majority-owned from an unsupported region is a fact that the company's shareholder records and its advisers must establish.

What to watch next

Three things are worth tracking. The first is whether Anthropic publishes how identity checks are carried out, since the policy page does not appear to describe the process in the reporting reviewed. The second is whether cloud platforms and resellers issue their own notices about Claude access in their regions. The third is whether the country list itself changes, because the May and November 2026 versions are reported as identical.

This is a company usage policy, not a statute. It sets out what Anthropic allows on its own services. This explains the rule in general terms; it is not legal or tax advice for your case.

How AiStaffo would automate this

AiStaffo would start with a map of every workflow that calls Claude, including the ones that run through a software vendor or a cloud account. For each workflow it would record the account holder, the countries where users sit, and the ownership details the owner provides. The system would flag any workflow that touches an unsupported region and hold a report for a person to review. The owner still decides whether to keep, move or pause each workflow, and confirms ownership with the company's own records. Book a free automation audit.

Questions people ask

Does the Anthropic Supported Regions Policy apply to Claude used through a cloud platform?
Yes. According to the policy as reported, it covers Claude models offered through cloud platforms and authorised resellers, not only the Claude.ai website and the API. A business that reaches Claude through a cloud account should check its own setup against the policy.
Does the policy apply if my company is outside China but has Chinese owners?
It can. The policy covers organisations directly or indirectly majority-owned or controlled from unsupported regions, wherever their staff are based. Whether that applies to a specific company depends on its ownership records, which an owner should review with the company's advisers.
What happens to an account that shows signs of use from an unsupported region?
According to reports on the policy, the account may face identity checks. Accounts that fail those checks are banned. A banned account stops any automated workflow that depends on it, so a fallback plan is needed.
Is Claude banned in China?
Mainland China is not on Anthropic's list of supported countries and regions, so the company does not offer Claude there. The policy also covers people in those places who use Claude while travelling or working remotely.
Did Anthropic add new countries to the ban this month?
Meduza reports that the country list is identical to the version dated May 2026. The November 2026 change mainly clarifies how the existing rules on location and ownership are enforced.
Do I need to change my automation tools now?
Not necessarily. A business should first check where its users sit, who owns the company and which vendors route work through Claude. The results decide whether any change is needed.

Book a free automation audit

Thirty minutes. We look at one process you run every week and tell you exactly what an AI worker would take off your desk, and what it would not.

We use these details only to reply to this request and prepare your audit, and keep them for up to two years. We send no marketing unless you tick the box above. Privacy policy

anthropicai regulationai governanceai modelsvendor risk