The U.S. Department of Homeland Security on Monday ordered civilian government agencies to remove DeepSeek — China's best-known AI chatbot — from their computer systems and to stop buying it, along with any outside service built on top of the model. The directive also bars agencies from making new purchases of DeepSeek products going forward.

The notice, available as a PDF via the Federal Register's public inspection system, warns that government or vendor use of DeepSeek "may result in the transfer of sensitive data to the PRC" — the People's Republic of China — and could produce "actionable intelligence on U.S. persons and/or government agencies." DHS says the models pose "multiple threat vectors," meaning several separate ways things can go wrong.

The ban reaches past DeepSeek's own app to third-party services developed using it, so coverage extends to any tool built on the model rather than the chatbot alone.

DeepSeek is an open-source model that can run locally or be hosted by any provider, not only China-based ones. Perplexity, for example, runs the model on its own servers located in the United States. The developers, however, host a version on servers located in China for those using Deepseek.com's services.

The order follows earlier restrictions across federal and state government. A July DHS memo, made public on Monday, had already told agencies to monitor traffic flowing in and out of government websites for DeepSeek-related risks. The Commerce Department and the Navy previously banned DeepSeek on government devices, and New York, Tennessee and Virginia have taken similar steps. The new directive extends that patchwork beyond device-level bans at individual departments and states, requiring removal from civilian agencies' systems and barring future purchases across them. The Chinese Embassy in Washington did not immediately comment.

DeepSeek, a Hangzhou-based startup, became a household name on Jan. 27, 2025, when its open-source model, built for under $5 million, knocked Nvidia's stock down 16% and erased about $600 billion of the chipmaker's value. U.S. tech as a whole lost an estimated $1 trillion that day, as DeepSeek proved as capable as America's best models while costing a fraction of the price — with the possibility of running for free on the appropriate infrastructure. That day turned the little-known startup into the focal point of U.S. scrutiny of Chinese AI.

Italy's data protection authority also ordered DeepSeek to block its chatbot days later, calling the company's answers to its privacy questions "totally insufficient." In April 2025, a House select committee on the Chinese Communist Party alleged that DeepSeek secretly sends American user data to the Chinese government. The same committee said DeepSeek may have used about 60,000 Nvidia chips, including 20,000 restricted under U.S. export controls — the rules that bar selling the most advanced AI chips to China.

DeepSeek has kept shipping since. The preview of its V4-Pro model, released in April, is priced about 98% below OpenAI's GPT-5.5 Pro and is free to download and modify under an MIT license. This week, the company released an updated version of its V4.1-Flash model aimed at efficiency, with big quality gains. The product momentum continues even as the restrictions around it widen.

The notice lands in the middle of a wider federal push against Chinese technology. The Federal Communications Commission votes on October 29 on barring all Chinese labs from testing electronics for U.S. use.

This report draws on original coverage by Decrypt.