DHS orders DeepSeek removed from federal civilian IT and barred from agency procurement

Published in the Federal Register on 20 October, FASCSA Exclusion and Removal Order 2026-01, signed by Homeland Security Secretary Markwayne Mullin and first issued 24 July, bars products and services of Hangzhou DeepSeek Artificial Intelligence Co. and its affiliates from civilian agency procurement. Agencies must also remove DeepSeek products, and third-party products built on DeepSeek, from their IT. National security systems are outside its scope.

Verified October 25, 20262 primary sourcesLean methodCorrections

By Daily Teabag Desk

Published Oct 25, 2026, 1:01 a.m. EDT

LeanCentre(0)Methodology

LeftCentreRight

Why this lean: Quotes the Federal Register notice and the order’s own terms and limits, especially that it covers civilian agencies only and is not a consumer ban. Scored 0.

ShareX / Twitter Bluesky LinkedIn Email

United States Capitol, west front — illustrative for a federal government order — Photo: Architect of the Capitol, public domain

The Department of Homeland Security published notice on 20 October 2026 (91 FR 66346, FR Doc. 2026-21447, Docket DHS-2026-0628) of an order under the Federal Acquisition Supply Chain Security Act (FASCSA). The order (1) excludes information technology products and services developed by Hangzhou DeepSeek Artificial Intelligence Co., Ltd and affiliated entities from federal civilian agency procurements, and (2) requires removal of all DeepSeek and affiliate products, “and third-party products and services developed using DeepSeek implementations”, from civilian agency information technology.

What the order says

Exclusion and Removal Order #2026-01, originally released 24 July 2026, signed by Secretary Markwayne Mullin. It is reproduced in the notice as images of the signed order.
Scope: federal civilian executive-branch agencies. National security systems are excluded from its scope.
Agencies may request exceptions within 60 days, and CISA reviews them. The order is reviewed annually.
The risk summary cites data stored on servers in the People’s Republic of China, collection of personal information, and security weaknesses found in DeepSeek’s iOS app.

… removing all DeepSeek and DeepSeek-affiliate products, and third-party products and services developed using DeepSeek implementations, from civilian agency information technology. — DHS notice, 91 FR 66346

Citizen impact: This governs federal computers and contracts, not private users. Contractors selling to civilian agencies need to make sure their products are not built on DeepSeek models, or seek an exception through an agency. Canadians: this is a U.S. federal procurement rule. Ottawa’s own departmental guidance is separate and is not covered here.

What this piece refuses

This is not a ban on Americans using DeepSeek, and we do not call it one. We do not repeat security claims about DeepSeek beyond what the order’s risk summary states. We have not reported company comment, because the notice contains none.

Bottom line

FASCSA order 2026-01 bars DeepSeek, its affiliates, and products built on DeepSeek implementations from federal civilian IT and procurement. National security systems are excluded, and agencies have 60 days to request exceptions.

United States Capitol, west front — illustrative for a U.S. Treasury rulemaking — Photo: Architect of the Capitol, public domain

Policy Watch|United States

Treasury proposes outright ban on U.S. investment in Chinese and other “country of concern” firms building any “SI system”

Treasury’s proposed rule implementing the Comprehensive Outbound Investment National Security (COINS) Act, issued 22 October and due in the Federal Register on 26 October, would make investment in country-of-concern entities that develop any “super intelligence (SI) system” a prohibited transaction, ending the current notification tier. It adds hypersonics, high-performance computing and biotech to the covered sectors, and adds a $50,000 de minimis exception. Comments close 7 December.

Daily Teabag Desk·October 25, 2026Centre

Berlaymont building, European Commission, Brussels — Photo: EmDee, CC BY-SA 4.0

Policy Watch|European Union

EU AI Act: a ban on AI that makes non-consensual sexual deepfakes applies from 2 December; high-risk rules are pushed to December 2027

The ‘Digital Omnibus on AI’, Regulation (EU) 2026/1744, has been in force since 27 July. It adds a new prohibition on AI practices that generate non-consensual sexual or intimate content and child sexual abuse material, applying from 2 December 2026, and sets the same date as the deadline for older generative systems to mark synthetic content. It fixes the delayed start of the high-risk rules at 2 December 2027 (listed uses such as hiring, schooling and biometrics) and 2 August 2028 (AI in regulated products).

Daily Teabag Desk·October 23, 2026Centre

James A. Byrne United States Courthouse, Philadelphia, seat of the Third Circuit — Photo: ajay_suresh, CC BY 2.0

Accountability|United States

First U.S. appeals ruling on AI training: copying Westlaw headnotes to train a rival legal search tool was not fair use

On 29 September the Third Circuit affirmed that ROSS Intelligence infringed Thomson Reuters’ copyright by using 2,243 Westlaw headnotes to train a non-generative AI legal search engine meant to compete with Westlaw. Three of the four fair-use factors weighed against ROSS, including harm to a developing market for licensing headnotes as AI training data. The court stressed that ROSS’s tool could not generate new text, leaving the generative-AI question open.