Trump administration bans China’s DeepSeek AI from civilian federal systems

The Trump administration has ordered civilian federal agencies to remove China’s DeepSeek AI from their systems over security risks, a step DHS warns could stop sensitive U.S. data from reaching Beijing.

Federal civilian agencies must strip out any DeepSeek-related technology and stop buying the Chinese AI firm’s products, the Department of Homeland Security disclosed this week. The order, issued July 24, covers departments including Treasury, State, and Justice. It does not reach the Department of Defense.

Breitbart News reported the directive also hits third-party products and services built on DeepSeek’s AI models. Agencies that need more time or an exception have 60 days from the order to send written justification to DHS.

DHS framed the move in plain security terms. The models and related technology, the department said, present clear risks to government systems and the data those systems hold.

DHS flags multiple threat vectors and data risk to China

The core worry is straightforward. Sensitive U.S. government information could end up in China if DeepSeek tools stay in federal hands. DHS also pointed to open-source research that found several critical vulnerabilities in DeepSeek’s iOS app.

The New York Post reported the order cites “multiple threat vectors” and warns use by government employees or vendors “may result in the transfer of sensitive data to the People’s Republic of China.” The action rests on the Federal Acquisition Supply Chain Security Act after a Federal Acquisition Security Council review. It builds on earlier Commerce Department and Navy restrictions.

That supply-chain focus matches a broader Trump administration push to keep hostile technology out of federal networks rather than try to manage it after the fact. The same D.C. apparatus that has faced fights over infrastructure and facilities projects is now being told to clean Chinese AI out of everyday civilian systems.

"present multiple threat vectors."

DHS used that language to describe DeepSeek’s models and technology. The department’s notice made the data-transfer risk explicit for anyone still tempted to treat the platform as just another software tool.

Beijing calls the ban politicized

China’s foreign ministry pushed back hard. Spokesperson Mao Ning, answering a Reuters inquiry, rejected the security framing and cast the order as politics dressed up as protection.

Mao Ning said China

"consistently opposes the over-generalization of national security as well as the politicization of science, technology and trade issues,"

Beijing also said it would safeguard the rights and interests of its companies. The statement offered no technical rebuttal of the vulnerability findings or the data-flow concern DHS put on the record.

The gap is familiar. U.S. agencies list concrete threat vectors and open-source flaws. Chinese officials answer with complaints about “politicization.” Taxpayers and federal workers left holding the systems get the risk either way.

Hall warned about CCP Trojan horses months earlier

Wynton Hall, Breitbart News social media director and author of the New York Times bestseller Code Red: The Left, the Right, China, and the Race to Control AI, had already flagged the danger. In a March interview with the Daily Mail, Hall urged that federal employees be blocked from the Chinese AI platform months before the July order landed.

He tied the DeepSeek problem to the larger pattern of Chinese tech that gains a foothold first and creates leverage later. Hall told the Daily Mail:

"The safest way to stop CCP Trojan horses is to not let them enter the gates to begin with,"

On TikTok, he added that “the jury is still out on whether the new U.S. deal’s implementation will end the myriad security risks that ignited the TikTok ban in the first place.” The same logic applies here: once the code is inside government networks, the damage is harder to reverse than a clean ban at the door.

Hall’s earlier call tracks with the administration’s final decision. Blocking the product from civilian agencies is the gate-keeping step he described. It also fits a White House record of treating security as an operational duty rather than a talking point, the same posture seen when Trump vowed jail time for vandals targeting major D.C. landmarks.

Civilian agencies only, for now

The July 24 order reaches Treasury, State, Justice, and other civilian departments. Defense stays outside this particular directive. Agencies that cannot meet the removal timeline, or that claim a unique need, must file written justification with DHS inside the 60-day window and request any compliance delay up front.

That process puts the burden on the agency that wants an exception, not on the security standard. It also keeps the default clear: DeepSeek-related technology comes out, and new purchases stop. Third-party tools that quietly sit on DeepSeek models fall under the same rule.

Federal workers and contractors now have a bright line. The alternative, leaving Chinese AI models inside systems that handle tax data, diplomatic cables, or justice records, is exactly the exposure DHS described. A capital already dealing with federal responses to vandalism at national memorials does not need another preventable vulnerability layered on top.

Prior Commerce and Navy limits showed the same direction. The new DHS order extends the scrub to the broader civilian side of government and locks in a purchase ban going forward. Messaging from the White House has stayed consistent on the China tech threat even as press shop personnel have shifted.

Security first, excuses second

DeepSeek is a Chinese AI startup whose models and apps drew official U.S. scrutiny for both technical holes and the destination of any data they might collect. DHS did not treat those as theoretical. It ordered the technology out of civilian federal IT and barred future buys.

China answered with the familiar charge that America is politicizing science and trade. The department’s own language stayed on the risks: multiple threat vectors, sensitive data that could move to the People’s Republic of China, and documented flaws in the company’s iOS app.

Readers who watch the capital know the pattern. Strong security steps draw immediate foreign protest and slow-walk pressure at home. The administration’s answer here was to issue the order, set the 60-day clock, and leave the exception process in writing. No speech replaces a clean system.

Keeping Chinese AI out of civilian government networks is basic hygiene, not a provocation, and the officials who treat it as optional are the ones gambling with everyone else’s data.