Why Anthropic Is Banning Claude for Companies Linked to China and Other Restricted Nations

On September 5, Anthropic, the developer behind the Claude large language model, issued an update to its sales restrictions, stating that companies with majority Chinese ownership can no longer use Claude, regardless of where they operate.

This marks the first time a U.S. AI firm has imposed such a policy. Any entity directly or indirectly owned more than 50% by a Chinese‑controlled group is barred from accessing Claude’s services, even through subsidiaries or cloud platforms.

The company cited legal, regulatory, and security risks as the primary reasons, noting concerns that these firms could use the technology to advance their own AI development and compete with U.S. and allied tech companies.

The restriction also extends to entities in Russia, Iran, North Korea and other countries the United States classifies as “adversary nations.” Direct customers and those accessing Anthropic via cloud services will be affected, with the company estimating a revenue impact in the hundreds of millions of dollars.

One insider indicated that the policy particularly targets Chinese subsidiaries established in Singapore that access U.S. technology with reduced scrutiny.

Anthropic’s full announcement emphasizes that its service terms now prohibit use by any organization whose ownership structure places it under the control of jurisdictions prohibited by its policy, regardless of the actual operating location. The statement also calls for stricter export controls, accelerated U.S. energy projects to support large‑scale AI infrastructure, and rigorous assessment of AI models for national‑security risks.

The company stresses that responsible AI firms must take decisive action to ensure transformative technologies serve democratic interests and do not fall into the hands of hostile actors.