Anthropic moves to close loopholes that allow Chinese access to Claude

Anthropic is moving to close workarounds that let Chinese companies reach Claude through offshore entities and cloud infrastructure, according to the Financial Times.

The reported enforcement push comes after Anthropic already decided in September 2025 to stop selling AI services to majority Chinese-owned groups, as well as organizations linked to Russia, Iran and North Korea. The new effort is aimed at making that policy harder to evade.

The report said some Chinese firms have used foreign subsidiaries and cloud services to access Claude even when they were supposed to be blocked under Anthropic policy. Those routes do not necessarily violate law in every jurisdiction, but they do appear to breach the company’s terms of service.

Claude has become especially valuable to engineers and companies interested in coding and model-distillation work, which helps explain why access remains attractive even as restrictions tighten. The reporting names ByteDance, Alibaba and Ant Financial among the Chinese groups discussed in the broader coverage.

How the workarounds reportedly worked

According to the Financial Times, some of the access has been routed through overseas subsidiaries, including Singapore-based entities, and through cloud infrastructure such as Microsoft Azure services.

Rather than relying only on static account checks, Anthropic is reportedly using dynamic detection methods to identify suspicious access patterns. Those methods include time-zone checks and traffic-pattern analysis meant to spot behavior that does not match the claimed account location.

The company is also said to be working with cloud partners, including Microsoft, to monitor and shut down unauthorized use. That suggests Anthropic is trying to police both the customer layer and the infrastructure layer that can be used to mask access.

Policy backdrop

This is not the first China-related restriction Anthropic has made public through reporting. The September 2025 move to stop selling to majority Chinese-owned groups was already framed as a national-security measure, and it also covered groups linked to Russia, Iran and North Korea.

The new report describes a further enforcement step rather than a reversal of that policy. In practice, Anthropic appears to be trying to turn a formal rule into one that is harder to sidestep through corporate structures or cloud routing.

That matters because offshore subsidiaries are a common way multinational businesses structure access to U.S. software and services. The reporting suggests Anthropic now views those structures as a loophole that needs active monitoring, not just a policy statement.

What to watch next

The main open question is whether Anthropic issues a formal public statement or policy update beyond the reporting. The Financial Times did not say which enforcement tools are already fully deployed and which are still being rolled out.

It is also unclear how broadly cloud providers will cooperate with Anthropic’s monitoring efforts. If the company expands verification, payment checks or network-based screening, similar access-control moves could spread to other frontier AI vendors.

For now, the story points to a broader shift in how AI companies are policing access in sensitive markets: not just by writing restrictions, but by tracing how customers actually reach the product.