The U.S. government has mandated that all federal civilian agencies cease purchasing and remove from their systems any products or services developed by the Chinese AI company DeepSeek. This directive, issued by the Department of Homeland Security on October 20, 2026, cites national security concerns linked to DeepSeek's operations in China. The prohibition covers not only DeepSeek’s own AI tools but also third-party products built using its technology, although agencies can seek waivers in exceptional cases.

The DHS’s primary worry stems from how DeepSeek handles user data, cautioning that sensitive information accessed by U.S. government personnel could be vulnerable to Chinese government oversight under China’s laws. This action follows a prior December 2025 law that limited DeepSeek’s involvement in intelligence and national security systems. The new order broadens restrictions to encompass all federal civilian agencies, addressing both existing technology and future acquisitions.

DeepSeek has been a significant player in AI competition between the U.S. and China, known for its cost-effective open-weight models that challenge U.S. AI giants like OpenAI and Meta. However, it has faced intense scrutiny internationally due to allegations of using "model distillation" techniques to extract information from American AI models. OpenAI and Anthropic have reported suspected attempts by DeepSeek to distill their proprietary AI outputs, raising further security alarms.

China has condemned the U.S. move, with Foreign Ministry spokesperson Mao Ning accusing Washington of politicizing technology and trade issues under the guise of national security. Meanwhile, China is contemplating export controls on several AI-related technologies including advanced models and chip designs, aiming to manage security risks. DeepSeek’s rapid rise since its R1 reasoning model debut in early 2025 has stirred significant geopolitical tensions over AI technology transfer and security worldwide.