The United States has formally barred DeepSeek from federal civilian government systems, in the broadest American restriction yet on the Chinese artificial intelligence company. The Department of Homeland Security made the order public on Monday, October 19, requiring civilian agencies to remove DeepSeek-related technology from their information technology systems and barring them from buying DeepSeek products and services in the future.
The ban reaches further than the company's own apps. It also covers third-party products and services developed using DeepSeek's technology or its AI models, which means a vendor tool built on a DeepSeek model is treated the same as DeepSeek software itself. That vendor clause is likely to be the hardest part for agencies to police, because an AI model can sit quietly inside another company's product.
The order was originally issued on July 24, months before it became public. It applies across federal civilian agencies, including the Treasury, State and Justice departments, but it excludes the Department of Defense. Agencies could request an exception, including a delay in complying, by submitting a written justification to the department within 60 days of the order being issued.
Homeland Security said DeepSeek's AI models present "multiple threat vectors." The order cited the risk that sensitive United States government data could be transferred to China if employees or vendors used the models, potentially giving Chinese authorities intelligence on American citizens and government programs. The department also warned that the technology could be used to track or digitally fingerprint users, and that the models may contain malware or other security vulnerabilities, pointing to open-source research that identified multiple critical vulnerabilities in DeepSeek's iOS app.
The legal basis is the Federal Acquisition Supply Chain Security Act of 2018, and the department said it acted following a review by the Federal Acquisition Security Council, the body that manages supply chain risks in federal procurement. In effect, Washington is treating a popular AI model the way it has previously treated risky networking hardware: as a supply chain problem, not just a software choice.
Beijing pushed back on Tuesday. Speaking at a regular press briefing, Chinese Foreign Ministry spokesperson Mao Ning criticized Washington for overstretching national security concepts and politicizing technological and economic issues, and said China will resolutely safeguard the legitimate rights and interests of Chinese enterprises.
The order is the culmination of a steady tightening that began soon after DeepSeek burst onto the scene at the start of 2025, when its low-cost R1 model stunned the industry by matching leading American systems on several complex tests. Commerce Department bureaus told staff in March 2025 that DeepSeek was banned on government devices, the Navy directed its members not to use the model, and Congress later restricted its use in intelligence community systems.
For federal technology vendors, the practical work starts now: auditing products for embedded DeepSeek components, documenting what they find, and being ready to show agencies that their tools are clear. For the wider AI industry, the order is another sign that where a model comes from, and where its data can go, now matters in government buying as much as how well it performs. DeepSeek did not immediately respond to a request for comment on the order.