The U.S. Department of Homeland Security has issued a directive requiring federal civilian agencies to cease the procurement of products and services from the Chinese AI startup DeepSeek. Agencies are also mandated to remove any existing DeepSeek-related technologies from their IT infrastructure due to national security concerns.
The restriction encompasses not only DeepSeek's proprietary offerings but also third-party solutions that integrate the company's technology. While the order is broad, federal agencies retain the right to request specific exceptions to these requirements.
The decision follows a December 2025 law that restricted DeepSeek's use within national security and intelligence systems. The new directive expands these limitations to the broader federal civilian sector, affecting both new acquisitions and current deployments.
U.S. authorities have cited risks regarding the handling of user data, suggesting that sensitive government information could be accessed by Chinese officials under local laws. Additionally, the U.S. has accused DeepSeek and five other firms, including Alibaba and Moonshot AI, of using model distillation to extract data from American AI systems.
OpenAI and Anthropic have reported observing unauthorized attempts by DeepSeek to distill their models. In response to the U.S. actions, China’s Foreign Ministry stated that Beijing opposes the politicization of technology and trade, pledging to protect the rights of its domestic enterprises.
