Claude AI Access in Mainland China: Rules and Risks

0

Anthropic's regional rules, its ownership-based limits and a new security investigation into unauthorized relay services all shape how Claude is reached from China.

AI-generated image showing Claude AI access restrictions in China, with a laptop, world map and security symbols.
Anthropic's regional access restrictions limit Claude availability in mainland China. (AI-generated image)

Claude is not an officially available product in mainland China. Anthropic's Supported Regions Policy limits access to the countries and regions it lists, and mainland China is not among them. A recent Symantec investigation into unauthorized relay services, plus a report about tighter enforcement, has put the issue back in the spotlight.

The picture has several separate pieces. Anthropic's geographic limits, its ownership-based restrictions, a reported November enforcement date and third-party access services are different things, and they are easy to confuse.

What Anthropic's Regional Policy Covers

According to Anthropic's policy, the restrictions apply across its product line. That includes Claude.ai, Claude Code, Claude Cowork, the company's API and developer platform, and Claude models offered through authorized cloud platforms and resellers.

The rules look at who is using Claude as well as where they are. They cover individuals physically located in unsupported regions. They also cover organizations incorporated or headquartered in those regions, and certain organizations that are majority-owned or controlled by entities based there.

Taiwan is listed separately as a supported region. Its status should not be treated as identical to that of mainland China.

Why Ownership Rules Were Added

On September 4, 2025, Anthropic announced stronger restrictions aimed at companies controlled by entities in unsupported jurisdictions, including China. The company said the change was meant to address legal, regulatory and national-security risks.

The ownership threshold is more than 50%. An organization owned, directly or indirectly, by companies headquartered in unsupported regions falls under the restriction. Anthropic's announcement means a subsidiary operating outside its parent company's home country can still be covered.

Anthropic gave several reasons for the policy:

  • Companies subject to certain foreign government requirements could potentially be compelled to share data or cooperate with intelligence services.
  • Advanced AI capabilities could be accessed without authorization.
  • Those capabilities could be used in military and intelligence applications.
  • AI models could be targeted for distillation.

Symantec Flags Unauthorized Relay Services

On October 29, 2026, Symantec's Threat Hunter Team published an investigation into unauthorized relay services that give access to OpenAI Codex and Anthropic's Claude Code. The researchers identified the activity across approximately 12 organizations. Most were in mainland China, with additional activity involving South Korea, Taiwan and Malaysia.

Symantec said the service installed genuine versions of the coding tools but altered their configurations. Requests were then routed through third-party infrastructure instead of the official vendor networks.

What the Configuration Changes Could Do

Symantec reported that the changes could:

  • Disable important execution sandbox protections.
  • Enable network access for the write sandbox.
  • Replace legitimate API credentials with credentials issued by the relay.
  • Turn off telemetry and error reporting.
  • Send prompts and files through third-party servers.

Symantec said it had not observed evidence that the relay operator actually abused its access. The investigation describes unauthorized access infrastructure and its security risks. It does not report a confirmed data-theft incident.

Anthropic's Own Threat Report

Separately, Anthropic's September 2026 threat intelligence report discussed suspected state-linked misuse of Claude. It covered activity between December 2025 and August 2026 and included a suspected China-based operation that Anthropic described as aligned with the Chinese government and focused on intelligence. Anthropic said it disrupted the activities described in the report.

The attribution is Anthropic's own assessment. It has not been presented as an independently proven conclusion.

Reported November Enforcement Date

The most recent development comes from TokenPost, which reported on November 11, 2026, that Anthropic's updated regional enforcement rules would take effect on November 12. The outlet also reported that Anthropic began rolling out identity verification checks in April, and that verification may involve government identification and a selfie submitted through Persona. According to TokenPost, unauthorized resale of Claude access and use from restricted regions could trigger additional verification, and failed verification could potentially lead to account closure.

Those details come from TokenPost alone. An original Anthropic announcement confirming the November 12 date, the October 8 announcement date or the verification specifics was not located. It also remains unclear whether every affected user would have to complete identity verification. No figures have been published for suspended accounts, terminated subscriptions or financial losses.

The November enforcement date is a reported detail and has not been confirmed by Anthropic.

What Is Confirmed and What Is Not

Several points are settled. Mainland China is not an officially supported Claude region, and that was true well before this month's reports. Anthropic has regional access restrictions, and it announced ownership-based limits in September 2025. Symantec has documented unauthorized relay services involving Claude Code and approximately 12 organizations.

Other points remain open. The exact enforcement date and the identity-verification process still need confirmation from Anthropic. It is also unknown whether the relay operators accessed or misused any confidential information. Readers should treat the November reporting as unconfirmed until Anthropic addresses it directly.

Anthropic's supported-regions policy and its September 2025 announcement are the primary sources for the company's restrictions, and they can be checked on its website.

Post a Comment

0 Comments
* Please Don't Spam Here. All the Comments are Reviewed by Admin.
Post a Comment (0)
To Top