Anthropic Updates Regional Sales Restrictions to Prevent Adversarial Access

Anthropic has updated its regional sales restrictions to prevent companies controlled by entities in unsupported regions—specifically citing adversarial nations like China—from accessing its AI services. This move is designed to mitigate national security risks associated with authoritarian regimes' legal requirements to share data and cooperate with intelligence services.

Regional Access Restrictions and Ownership Structures

Anthropic is prohibiting access to its services for companies or organizations whose ownership structures subject them to control from jurisdictions where Anthropic products are not permitted. This restriction applies regardless of the following:

• Operating Location: The prohibition applies even if the company operates in a supported region.
• Ownership Threshold: The restriction targets entities that are more than 50% owned, directly or indirectly, by companies headquartered in unsupported regions.

This update is intended to close loopholes where companies from restricted regions accessed services through subsidiaries incorporated in other countries.

National Security and Strategic Risks

Anthropic identifies several risks associated with providing AI capabilities to entities controlled by authoritarian regions:

• Intelligence Cooperation: Legal requirements in authoritarian regions can compel companies to share data or cooperate with intelligence services, creating national security risks.
• Military and Intelligence Applications: Access to Anthropic's services could be used to develop applications that serve adversarial military and intelligence services.
• Sourcing and Distillation: Adversarial entities could use these models to advance their own AI development through techniques like distillation, allowing them to compete globally with technology companies in the United States and allied countries.

Policy Advocacy and AI Safety

To support these restrictions, Anthropic continues to advocate for broader systemic policies to prevent the misuse of transformative AI by authoritarian adversaries. These include:

• Export Controls: Supporting strong export controls to prevent authoritarian nations from developing frontier AI capabilities.
• Infrastructure Investment: Accelerating energy projects on US soil to build out the large-scale infrastructure necessary for AI scaling.
• Rigorously Evaluating Models: Conducting rigorous evaluations of AI models for capabilities that could be exploited by US adversaries.

Anthropic states that the safety and security of AI development requires a collective commitment to ensuring that transformative technologies serve US and allied strategic interests and support democratic values.