By Hamza Faisal
The United States has drawn its hardest line yet against Chinese artificial intelligence. The Department of Homeland Security on Monday ordered all federal civilian agencies to remove DeepSeek's AI products — including third-party services built on the Chinese lab's models — from their IT systems, effective immediately.
The directive also bars civilian agencies from acquiring DeepSeek products or any third-party services developed using its AI implementations going forward. DHS said DeepSeek's models pose "multiple threat vectors" and raised concerns that their use "may result in the transfer of sensitive data to the PRC if used by US government employees or vendors."
The order was published under docket DHS-2026-0628 and is due to appear in the Federal Register on Tuesday, October 20. It follows a recommendation from the Federal Acquisition Security Council and is issued under the Federal Acquisition Supply Chain Security Act of 2018, the same authority Washington has used against Kaspersky, Huawei and ZTE.
What the order actually does
The order tells civilian executive-branch agencies to do two things. First, exclude Hangzhou DeepSeek Artificial Intelligence Co., Ltd., its affiliates and their products from purchasing "effective immediately", including agency-approved product lists and e-commerce marketplaces. Second, remove DeepSeek products and "third-party products and services developed using DeepSeek implementations" from agency IT systems immediately.
That second clause is the wide one. DeepSeek's models are released as open weights, which means they turn up inside other companies' tools — not just the company's own app and website.
The ban does not reach the military or spy agencies — it applies to federal civilian systems "except for their national security systems and secure compartmented information systems", which fall under the Pentagon and the Director of National Intelligence. It also does not, for now, extend to the private sector, add DeepSeek to the Commerce Department's Entity List, or impose broader export controls.
Why DeepSeek, and why now
DeepSeek rattled the AI world in early 2025 when it released models that appeared to match leading American systems at a fraction of the development cost, briefly wiping billions off the market value of Nvidia and other chip stocks. Its low-cost open-weight models have kept challenging US developers on price and performance ever since.
The new order lands just days after DeepSeek updated its cheap Flash model, which the company says now beats Claude Opus 5 at some coding tasks. Several US agencies — including the Commerce Department and the Navy — plus multiple states had already imposed narrower restrictions. The DHS directive now extends the pressure across the entire civilian federal enterprise and any contractors that rely on government systems.
News of the ban broke on a day when Wall Street was already healing: the Nasdaq rose 1.75% as markets recovered from last week's AI-linked 4.5% sell-off.
What comes next for the AI decoupling
The move is the clearest sign yet that foundation models are becoming a front line in the US–China technology split, not just chips and telecom gear. By targeting third-party tools that embed DeepSeek's weights, Washington is forcing a rapid audit of the AI supply chain — every startup and vendor that touches government data or contracts will now face sharper questions about which models sit inside their products.
DeepSeek has not publicly responded to the order. For the broader industry, the message is unmistakable: in federal procurement, a model's nationality now matters as much as its benchmark scores.
By Hamza Faisal
