US Bans China’s DeepSeek From Federal Agency Systems

The DeepSeek federal ban is now official. The U.S. Department of Homeland Security has ordered all civilian federal agencies to remove products and services built on China’s DeepSeek artificial intelligence models from their IT systems, and to stop buying anything developed by the Chinese company going forward.

The order, published Monday in the Federal Register as document 2026-21447, cites “multiple threat vectors” from DeepSeek’s models. DHS said use of the products “may result in the transfer of sensitive data to the PRC if used by U.S. government employees or vendors,” according to Reuters’ reporting on the notice.

What the DeepSeek Federal Ban Covers

The directive — Exclusion and Removal Order #2026-01, signed by Homeland Security Secretary Markwayne Mullin — does two things at once:

• Removal: agencies must take all DeepSeek and DeepSeek-affiliate products, plus any third-party products and services developed using DeepSeek models, off their information systems
• Exclusion: agencies are barred from purchasing information technology products and services developed by Hangzhou DeepSeek Artificial Intelligence Co., Ltd. and its affiliates

The order rests on the Federal Acquisition Supply Chain Security Act of 2018 and applies across the federal civilian executive branch. National security systems and secure compartmented information systems are excluded, since those fall under Pentagon and intelligence community authorities.

Why Washington Says DeepSeek Is a Risk

DHS’s stated reasons center on data exposure. The order warns that DeepSeek’s servers operate in China, where domestic law gives government cyber and intelligence functions broad authority to access private company infrastructure. Prompts entered into DeepSeek applications, the department argued, could send sensitive data to China — and combined with public records, that information “could produce actionable intelligence on U.S. persons and/or government programs.”

The department also said DeepSeek’s models “themselves may contain malware or other security vulnerabilities,” and cited concerns about the platform tracking or digitally fingerprinting users.

Third-Party Tools Are Caught Too

The ban’s widest reach may be its second limb: it covers not just DeepSeek’s own app and models, but any contractor product built on them. A vendor that embedded a DeepSeek model inside a tool sold to the Treasury Department is now selling an excluded product — and must swap the model out or be shut out of civilian agency purchases.

Agencies may request written exemptions within 60 days through the Cybersecurity and Infrastructure Security Agency for mission-critical functions. The order’s cover page lists an original release date of July 24, 2026 — meaning the ban has effectively been in force for months, though the paperwork only became public this week.

A Crackdown Years in the Making

The DeepSeek federal ban caps a crackdown that began when the company’s R1 model stunned the tech industry. Trained on less powerful chips at a fraction of the cost of American rivals, the open-source system outperformed models from OpenAI and Anthropic on several complex benchmarks — and its free app was downloaded by millions of Americans.

Concerns followed quickly: cybersecurity researchers questioned DeepSeek’s data practices, and last year a full U.S. ban on the consumer app was under discussion. The Commerce Department, the Navy, and states including New York, Tennessee, and Virginia had already barred DeepSeek from their devices before the federal order.

What It Means for the AI Race

The order is the latest sign that Washington is treating frontier AI models as strategic assets — and Chinese ones as supply-chain risks. By extending the exclusion to third-party products built on DeepSeek, DHS is effectively pushing federal vendors toward American or allied AI providers.

The move lands amid intensifying technological rivalry with Beijing, from semiconductor export controls to the race over AI infrastructure. For agencies, the immediate work is practical: audit IT systems, remove covered products, tighten network logging, and monitor outbound traffic while the removal is underway. For the broader market, it is another signal that the U.S. government intends to keep Chinese AI out of its digital infrastructure — wherever it can.