Anthropic bans Claude for Russian and Chinese companies and their employees starting November 12

It used to be simple to decide whether you could use Claude: you just looked at which country you were sitting in. From November 12, things get more complicated. Anthropic is updating its supported regions policy, and now it is not only your passport or geolocation that matters, but also your employer's legal affiliation.

What exactly has changed

The policy applies across all product surfaces: Claude.ai, Claude Code, Cowork, the API, and access to the models through cloud platforms. The last point matters: the workaround of "I'll get Claude not directly, but through a cloud provider" formally does not exempt you from the rules.

Use is prohibited in three cases.

First: the person is physically located in an unsupported country. This is the most straightforward rule, and it was the main one before as well.

Second: the company is registered or headquartered in an unsupported country. In that case the restriction extends to its employees regardless of where they are. A developer from a Russian company who has moved to Georgia or Kazakhstan and works from there is still covered.

Third: the company is more than half owned by entities from unsupported regions, or is controlled by them, directly or indirectly. Here, too, the restriction follows the staff. Anthropic looks at registration, ownership structure, and corporate control, not just the office address.

Russia and China are not on the list of supported regions.

What this means in practice

Imagine a startup with a legal entity in Cyprus but with Chinese investors who hold 60% of the shares. The team sits in Lisbon, and every developer has Claude Code installed. Under the third rule, such a company formally falls under the ban, even though not a single employee is in China. The situation is similar for the foreign subsidiary of a Russian company: you can change the geography of the office, but you can't quickly change the control structure.

There is a separate risk for outsourcing firms and studios. If you are building a project for a client and running code through Claude, the question of "whose account is this and whose organization is it" becomes a legal question, not a technical one.

What has not been said

An important caveat from the policy itself: it establishes a ban on use, but does not guarantee an automatic block of every account. How exactly Anthropic will detect violations, how strictly it will check ownership structures, and what will happen to existing subscriptions and API keys is not disclosed in the post. There is no reason either to panic or to count on getting away with it. This is the company's decision, and it may enforce it selectively or strictly.

What to do

First, read the original: https://www.anthropic.com/legal/supported-regions. Retellings in Telegram channels, including this one, are no substitute for the text of the policy.

Then go through three questions: where your people are located, where the legal entity is registered, and who owns more than half of the company. If the answer to any of them is questionable, that is a reason to consult a lawyer rather than guess.

And the most important thing to do regardless of the answers: remove the single point of failure. If Claude is built into your product, your CI, or your team's workflow, keep a fallback. It could be another model behind the same abstraction layer, or an open model you can run yourself. A vendor's policy can change within a few weeks, and migrating on the fly is more expensive than doing it in advance.

Bottom line: Anthropic is starting to look at who stands behind an account. For teams with Russian and Chinese roots, including those that relocated long ago, this is no longer just a question of the country of residence, but of the structure of the business.