The United States has taken its strongest action yet against Chinese artificial intelligence. The U.S. Department of Homeland Security has ordered all civilian federal agencies to remove technology from the Chinese company DeepSeek from their computer systems and stop purchasing it. The ban targets not only the chatbot itself, but also software from other companies that relies on DeepSeek models. Washington fears that sensitive data belonging to U.S. officials could end up in the hands of Chinese authorities.
Homeland Security Secretary Markwayne Mullin signed the directive on July 24. However, the public did not learn about it until October 19, when the department published the notice. The secretary invoked a 2018 law that allows the government to remove technologies considered a supply chain security risk from its systems. The decision followed a recommendation from a special board overseeing the security of government procurement. Washington previously took similar action against the Russian antivirus software Kaspersky, for example.
Who the ban applies to and who it does not
The measure applies to all civilian federal agencies. These include the Departments of the Treasury, State, and Justice, as well as agencies such as the Environmental Protection Agency and the Federal Communications Commission, which oversees telecommunications. The Department of Defense is exempt. The ban also does not apply to national security systems and classified networks that fall under the Pentagon and the Director of National Intelligence. Nor does it affect ordinary Americans or private companies, which may continue to use DeepSeek.
However, software vendors that sell to the government will face additional concerns. Each of them must now verify whether a DeepSeek model is running anywhere in their product. Companies whose contracts contain the relevant clause must monitor new directives on the government portal and report if they have supplied agencies with such a product. Critics point out that the text does not explain precisely what is meant by software developed using DeepSeek. Nor does it say how to treat a DeepSeek model running on the servers of an American company.
Why Americans object to DeepSeek
The department claims that DeepSeek models threaten government systems in several ways. The company stores data on servers in China, where local laws allow authorities to access the systems of private companies. According to Washington, information entered by federal employees or contractors could therefore reach the Chinese government.
Mullin also warned that DeepSeek collects users’ personal data when they create an account, can track them, and can create a digital fingerprint that can be used to identify an individual. According to the department, the software itself may contain malicious code or other security vulnerabilities.
Also at issue is a dispute over so-called distillation, a process in which one model learns from the responses of another. Last month, the U.S. National Security Agency, the Cybersecurity and Infrastructure Security Agency (CISA), and the FBI accused several Chinese companies in a joint letter of conducting distillation on an industrial scale and named DeepSeek directly. A September report then accused DeepSeek and five other Chinese companies, including Alibaba and Moonshot AI, of extensively extracting knowledge from U.S. technologies. Beijing rejects such claims.
Gradually tightening the screws
DeepSeek attracted attention in early 2025, and Washington began taking action against it almost immediately. As early as January 2025, the U.S. Navy prohibited its personnel from using it, and the Defense Information Systems Agency blocked access to the website on Pentagon networks. In March, employees under the Department of Commerce were instructed not to use the technology on work devices. NASA took similar action.
Members of Congress Josh Gottheimer and Darin LaHood introduced legislation in February 2025 that would ban the application on all government devices. However, the bill never made it out of committee. In December 2025, the defense budget law for fiscal year 2026 took effect, banning DeepSeek from national security systems.
Individual states also acted independently. Texas became the first to ban DeepSeek on government devices in January 2025, followed by New York, Virginia, and Tennessee. The new directive thus replaces the existing patchwork of partial bans with a uniform rule for all civilian agencies.
Beijing is furious
Chinese Foreign Ministry spokesperson Mao Ning accused the United States of abusing the concept of national security and politicizing technology and trade. According to her, Beijing will firmly defend the legitimate interests of Chinese companies. However, she has not yet announced any specific countermeasures, and DeepSeek itself has not commented on the ban.
George Chen of the Washington consulting firm The Asia Group pointed out that the dispute demonstrates the deep strategic distrust between the two powers. This is despite President Donald Trump and Chinese leader Xi Jinping seeking to maintain a warm and constructive personal relationship. Tensions are also being fueled by U.S. restrictions that prevent China from accessing the most powerful artificial intelligence chips and the machinery needed to manufacture them. Last week, Beijing introduced its own proposal that would restrict exports of certain advanced artificial intelligence models.
DeepSeek may not be the last. A draft 2027 defense budget prepared by a congressional committee would expand the Pentagon’s ban to ten additional Chinese artificial intelligence companies and software derived from their products. According to available information, the White House has also considered restrictions on the Kimi chatbot from Moonshot.
Sources: wsj.com, cyberscoop.com and cnbc.com
