Anthropic Shuts Claude to Russian-Controlled Companies from November 12

For many teams, Claude stopped being a toy for experiments long ago. It is Claude Code in the editor, the API in production, the model inside someone else's SaaS product. Now, for some of these teams, the door is closing, and more broadly than the headlines might suggest.

What exactly has changed

From November 12, Anthropic is introducing an updated Supported Regions policy. Access to Claude is blocked for organizations that are registered in Russia or controlled by Russian persons. The restriction applies to every surface of the product: the Claude.ai web chat, Claude Code, the API, and models sold through cloud platforms and resellers.

An important detail: the deciding factor is neither an employee's passport nor the city where they sit. The rule is tied to the organization and to whoever owns it. A developer of any citizenship working at a company under Russian control formally falls under the restriction just like their colleagues.

Why "registering in another country" doesn't solve the problem

The key change concerns corporate control. The ban extends to companies that are directly or indirectly under majority control of persons from unsupported jurisdictions, even if the legal entity is incorporated in a third country.

A simple example. A startup is registered in Dubai or Delaware, the team is scattered around the world, but more than half of the shares belong to founders from Russia. Under the text of the new rules, such a company may also fall within the restricted zone. Previously, such arrangements often lived in a gray area. Now Anthropic explicitly states that it looks at the real owners, not the address in the charter.

According to the documents, the company presents this as a clarification of rules that were already in force, not as a new ban. It names its motive openly: preparing for an IPO and the need to clean up revenue from customers who violate the regional policy. That cleanliness matters to investors and auditors, so a softening anytime soon is unlikely.

What developers and product managers should do

There is no need to panic, but an audit is worth doing right now.

• Find out whether the rules apply to you. Look at your ownership structure: who controls the company and through which legal entities. Borderline cases are better discussed with a lawyer than guessed at.

• Map your dependencies on Claude. It is not only your direct API key. Claude may sit in a cloud marketplace, at a reseller, in third-party tools and in contractors' services. The restrictions are declared for these channels as well.

• Don't tie yourself to a single model. If you don't have an abstraction layer over the LLM provider, now is the time to build one: a unified calling interface, a set of regression prompts, quality evaluation on your own data. Then swapping a model becomes a task of a couple of weeks rather than a rewrite of the product.

• Check your suppliers. If you buy a service that uses Claude internally, ask the vendor how the new policy will affect you and your deal.

• Evaluate alternatives in advance. Other closed models and open-weight solutions are not identical in coding quality and agent performance. It is better to compare them on real tasks today than on the day of the shutoff.

Limitations and what remains unclear

This report is based on Anthropic's public documents as retold by Habr. Practical questions remain: how the company will verify owners, how strictly the rule will be applied to minority stakes and to those arriving through resellers, and how cloud platforms will behave. Only practice after November 12 will provide the answer.

The overall takeaway is broader than this specific news item. Access to frontier models depends on jurisdiction and ownership structure, which means it is a regulatory risk that must be built into your architecture and business plan alongside prices and limits. Teams that can already switch providers without drama will come out ahead whenever the rules change next.