US Government Orders Full Removal of Chinese AI DeepSeek from Civilian Federal Systems
-
Post By
Emmie
- October 20, 2026
The U.S. Department of Homeland Security (DHS) has officially ordered the removal and prohibition of all artificial intelligence technology produced by Chinese firm DeepSeek across federal civilian government agencies.
The directive, which was issued in July but made public on Monday, bans civilian entities including the Department of the Treasury, Department of State, and Department of Justice from acquiring or utilizing DeepSeek products, as well as any third-party services developed using its AI models. The Department of Defense is excluded from this specific civilian order, having previously implemented separate restrictions.
In its directive, the DHS warned that DeepSeek's open-source models present "multiple threat vectors," and expressed concerns that potential foreign intelligence collection "may result in the transfer of sensitive data to the PRC if used by US government employees or vendors."
They also said that the use of DeepSeek "could produce actionable intelligence on US persons and/or government agencies. The models themselves may contain malware or other security vulnerabilities."
Agencies were instructed to monitor web traffic flows to mitigate security risks, with a 60-day window provided to submit written justifications for temporary compliance exemptions.
Responding to the U.S. ban, China's Ministry of Foreign Affairs stated it would defend the commercial interests of domestic technology developers. Chinese Foreign Ministry spokesperson Mao Ning told reporters that "China consistently opposes the over-generalization of national security as well as the politicization of science, technology and trade issues."
The DHS order builds on previous restrictions enacted by Congress, the Commerce Department, the U.S. Navy, and several state governments including New York, Tennessee, and Virginia. It also aligns with wider Trump administration policy actions targeting Chinese hardware, telecommunications, and testing laboratories.
Within the U.S. tech sector, senior officials have raised intellectual property concerns regarding Chinese AI development. Earlier this year, White House tech lead Michael Kratsios stated in an internal memo that "malicious exploitation" of American AI frameworks had been occurring, primarily driven by competing firms based in China. American AI developers, including OpenAI and Anthropic, have similarly accused Chinese entities of copying model architectures.
DeepSeek gained global prominence following its initial public release in 2025, offering open-source AI models that perform competitively with leading American systems at a fraction of the development cost. While OpenAI spent roughly $100 million to train GPT-4, DeepSeek developed its debut model for approximately $6 million (£4.5 million), a cost gap that previously triggered notable declines in international tech stocks.
Because DeepSeek offers lower operational expenses compared to closed U.S. models like Google's Gemini, Anthropic's Claude, or OpenAI's Codex, numerous software developers and enterprise platforms including Snowflake, Vercel, and OpenRouter have incorporated DeepSeek models into commercial customer offerings.
Under the new DHS directive, federal civilian agencies must now audit their IT infrastructure and software supply chains to identify and eliminate any integrated third-party applications relying on DeepSeek's underlying code.
