A familiar AI name has returned to Washington's security agenda, but the paperwork tells a more precise story than a sweeping claim that DeepSeek has simply been banned across America. The new public filing is worth reading for both its reach and its dates.
The Department of Homeland Security notice made available on October 19, 2026 covers removal of DeepSeek products from federal civilian agency systems and exclusion from their procurement. Reuters reported the development on Monday. The underlying order, attached to the notice, gives its original release date as July 24, 2026; the notice is scheduled for Federal Register publication on October 20. AsiaNewsIran reviewed the document to separate those dates from the practical scope.
What the order covers
The requirements extend beyond an employee downloading the DeepSeek app. They cover the company, affiliated entities and third-party products and services developed using DeepSeek implementations. That makes the technology inside a supplier's product relevant, as well as the name printed on the product.
- Covered products are excluded from civilian agency procurement.
- Covered products and services are to be removed from civilian agency information technology systems.
- The order provides a process for agencies to request exceptions.
This is a federal supply-chain security action. Its stated scope is civilian executive-branch agencies, with national-security and sensitive compartmented systems subject to separate authorities. It is not a general prohibition on every American consumer or private company using an AI chatbot.
Why July and October both appear
The appendix identifies the measure as Exclusion and Removal Order #2026-01 and prints July 24 as its original release date. The Federal Register filing stamp is October 19, with publication scheduled for the following day. The fresh development is the public notice and accompanying disclosure; it would be misleading to describe the attached order as first issued on October 19.
The order calls for procurement exclusion and removal immediately after issuance, or an exception request. It says written exception requests must be submitted within 60 days of issuance. Readers should therefore not treat the October filing as the start of a new 60-day window. The notice does not establish how many agencies have already completed removal or how many exceptions have been granted.
The security case, and its limits
DHS points to risks involving data supplied to hosted models, servers in China, access to company infrastructure under Chinese law and security vulnerabilities. These are the department's stated reasons for the order. They should not be rewritten as proof that every user has been spied on or that every deployment sends identical information to the same destination.
DeepSeek's published privacy policy identifies its Chinese data controller and describes processing and storage in China. That provides relevant context, but it is not a response to this particular order. A specific new company response was not established in the sources reviewed for this report.
Protecting sensitive government information requires knowing what a system contains, where its data goes and who can access it.
That is the editorial case for taking supply-chain scrutiny seriously. It also calls for precision: a restriction on government procurement is different from a nationwide consumer ban, and a department's risk assessment is different from a finding about every individual user.
What readers should watch next
Three practical questions remain:
- How agencies document removal of affected products.
- How third-party services identify the models embedded in their systems.
- What exceptions and implementation guidance become publicly available.
The wider question of who can access collected information also runs through our report on MapQuest's surveillance-camera layer and the databases behind it. A visible interface rarely tells the whole data story.
The clearest reading of this filing is a targeted federal-agency restriction with an earlier order date, not a new blanket ban on private use. Which matters most to you when choosing an AI service: its capabilities, its data location or the rules governing access?