On September 4, 2025 Anthropic updated its Terms of Service to expand restrictions on companies and organizations whose ownership structures subject them to control from jurisdictions where Anthropic’s products are not permitted. The announcement specifically cites connections to authoritarian states such as China.
Anthropic says that legal and regulatory requirements in certain countries can compel companies under their influence to share data, cooperate with intelligence services, or take other actions that pose national security risks. These pressures, the company argues, can persist regardless of where a company operates or the personal views of its employees.
The company notes that despite existing regional prohibitions, entities tied to restricted jurisdictions continue to access services in various ways, for example through subsidiaries incorporated in other countries. When such entities obtain access to Anthropic’s models, they could:
- use the capabilities to build applications and services that ultimately serve adversarial military or intelligence services and broader authoritarian objectives;
- leverage the models to advance their own AI development through techniques such as distillation, potentially competing with trusted technology companies headquartered in the United States and allied countries.
What the update changes
The revised policy bars access to Anthropic’s products by companies or organizations whose ownership structure places them under control from jurisdictions where the products are already prohibited. The prohibition applies to entities that are more than 50% owned, directly or indirectly, by companies headquartered in unsupported regions. Anthropic says this adjustment aligns its Terms of Service with real-world risks and with the intent of its policies.
Broader policy stance
Anthropic reiterates its support for measures such as strong export controls to prevent authoritarian nations from developing frontier AI capabilities that could threaten national security. The company also advocates accelerating large-scale energy and infrastructure projects on U.S. soil to support AI scaling, and for rigorous evaluation of AI models for capabilities that are relevant to national security and could be exploited by adversaries.
Anthropic frames the safety and security of AI development as a collective responsibility to prevent misuse by authoritarian actors, and argues that responsible AI companies should take decisive action to ensure transformative technologies serve U.S. and allied strategic interests and democratic values.
