Anthropic has put a rewritten set of rules on who may use Claude into force, and the ownership test is the sharp end: any company majority-owned or controlled by firms in unsupported regions loses access, even if its staff sit in a country where Claude is sold. The Supported Regions Policy, effective 12 November, covers Claude.ai, Claude Code, Cowork, the API and models sold through cloud platforms — and leaves mainland China, Hong Kong, Russia and Iran off its list of roughly 190 supported places.
Three tests, not one
The policy turns on three tests, and each one reaches further than the last. First, nobody may use Claude while physically located in a region the company does not support — including the staff of companies Anthropic does serve elsewhere.
Second, entities incorporated or headquartered in an unsupported region are barred, along with their users and personnel, wherever those people happen to sit. Third, the same bar applies to any entity majority-owned or controlled, directly or indirectly, by persons or entities in unsupported regions.
The list itself covers roughly 190 countries and regions, and territories of a listed country count as supported — Puerto Rico, for example, as a territory of the United States. Anything not on the list is unsupported. Taiwan is listed. Ukraine is listed except for Crimea, Donetsk, Kherson, Luhansk and Zaporizhzhia.
The rules reach beyond Anthropic’s own apps. They cover the API and developer platform, and Claude models sold through cloud platforms and authorised resellers — so a company barred from claude.ai cannot simply buy the same model through a cloud provider instead. That breadth is what makes the ownership test bite: it follows the customer, not the storefront.

Hong Kong, again
The most consequential absence is Hong Kong. The territory is not on Anthropic’s supported list, which means companies headquartered there — and companies majority-owned from there — cannot use Claude under the new rules. That is the same territory whose investors are being turned away from Anthropic’s share sale by its banks, under export-control rules, days before the offering prices.
The timing is not an accident of the calendar. The rewritten policy took effect on 12 November, in the same week Anthropic’s roadshow drew orders worth about six times the shares on offer. The ownership test is not new — Anthropic has applied a version since September 2025 — but the rewrite states it plainly.
Our Read
Read together, the policy and the IPO exclusion tell one story: Anthropic is drawing a hard line between its business and China, and it is doing so before it becomes a public company rather than after. The product rule and the share-sale rule now point the same way.
A usage policy is easier to defend than a share allocation, because it applies to a product Anthropic controls outright. But it also costs the company something real: developers and customers in the excluded regions it will now turn away, and subsidiaries abroad whose ownership alone disqualifies them.
Why It Matters
Claude is now one of the most widely used AI systems in software development, and Anthropic’s access rules are becoming a template other labs watch. When a frontier lab bars whole classes of companies by ownership, it changes who can build on the leading models.
It also signals how export controls are now enforced: not only at the border, but through the terms of service of the companies that sell the models. For any firm with Chinese ownership weighing a Claude deployment, the question is no longer whether the rules apply — it is whether the company itself is on the wrong side of them.
Background
Anthropic announced the rewrite on 8 October, and the new version took effect on 12 November. The company is due to price its shares on 18 November, with trading expected the next day. The ownership test is the part most likely to be challenged first, as companies check whether their shareholders put them on the wrong side of the line.


0 Comments