A list of 185, and everything outside it
Anthropic’s Supported Regions Policy took effect on 12 November. It names 185 countries and regions where the company offers its products, and states plainly that “any country or region not listed is unsupported”.
The policy covers the whole product line — the Claude.ai, Claude Code and Cowork apps, the API and developer platform, and Claude models offered through cloud platforms and authorised resellers. Territories and dependencies of a listed country are supported too, with Puerto Rico given as the worked example.
Mainland China, Hong Kong and Macau are not on the list. Neither are Russia, Belarus, Iran, North Korea, Cuba, Syria, Afghanistan, Myanmar, Venezuela or Yemen. Taiwan is listed. Ukraine is listed with an exception carved out for the Crimea, Donetsk, Kherson, Luhansk and Zaporizhzhia regions.

The part that reaches past borders
The geography is the easy half. The harder half is the three-pronged definition of what counts as unsupported use, which follows corporate structure rather than location alone.
The first prong is conventional: people physically in an unsupported region may not use the products, and that includes the personnel of otherwise supported entities. An employee of a supported company cannot use Claude while sitting in an unsupported country.
The second extends to incorporation. Entities incorporated or headquartered in an unsupported region are out, and so are their users and personnel, “regardless of whether such individuals are physically located in a supported region”.
The third is the one with the widest reach. Entities “majority-owned or controlled, directly or indirectly, by persons or entities in unsupported regions” are also unsupported, again including their users and personnel wherever those people sit. A subsidiary registered in a supported country, staffed by people in supported countries, falls outside the policy if control traces back to a region that is not on the list.

Enforced through the usage policy
The regions policy does not stand alone. Anthropic’s Usage Policy, which carries the same 12 November effective date, lists among its universal prohibitions accessing, facilitating or providing others — including end users — with access to the products “in violation of our Supported Regions Policy”.
That places a region breach in the same category as the policy’s other universal standards, and brings it inside the enforcement mechanism described there: Anthropic says its Safeguards team implements detection and monitoring, and that where it suspects a violation it may warn a user or throttle, limit, suspend or terminate access.
Why the control clause matters
The indirect-control language is what makes this more than a geoblock. Geoblocking is defeated by a virtual private network; a corporate-control test is defeated only by changing who owns the company. For enterprises with cross-border ownership structures, the obligation effectively shifts from knowing where staff are to knowing who controls the entity employing them, and demonstrating it.
It also lands at a particular moment for Anthropic. The company priced its proposed initial public offering at $70 to $78 a share on 10 November, two days before these policies took effect. A company preparing to list has reason to put its access restrictions in writing with precision.
What to watch is enforcement rather than text. The policy sets out who is excluded; it does not describe how ownership and control will be verified, or what a customer must produce to show that an unsupported region does not sit somewhere in its ownership chain.