The US DeepSeek Order Reaches Beyond the Chatbot. Here Is Its Scope.
![]() |
| AI-generated illustration of government technology procurement; not a photograph of a federal agency. |
When I see a headline saying a country has banned an AI model, the first thing I look for is the scope. Does it concern public procurement, a particular agency, company devices or everyone’s personal phone? Those are very different stories.
The new public notice concerning DeepSeek is about US federal civilian systems and procurement. Its most interesting detail is that the restriction reaches beyond an app with DeepSeek’s name on the icon.
Who the order covers
The Federal Register notice published on October 20 contains an order originally issued on July 24, 2026. It directs Federal Civilian Executive Branch agencies to exclude covered DeepSeek products and services from procurement and remove them from their IT systems.
The signed order also covers DeepSeek affiliates and third-party products and services developed using DeepSeek implementations. It excludes the agencies’ national security and secure compartmented information systems from this particular order’s scope.
The required actions are immediate from issuance. The text provides an exception process, with written requests due within 60 days of issuance. The October publication does not create a fresh 60-day grace period.
This is not an order banning ordinary consumers throughout the United States from using DeepSeek.
The name on the app is only the beginning
That third-party provision is what makes the story relevant beyond a single chatbot. An organisation can buy a writing assistant, a support tool or a coding service without the underlying model being obvious from the product’s branding.
For a covered agency, checking the installed app list alone would be an incomplete way to understand exposure. The supplier relationship and the technology used behind the service matter too.
I also would not assume that hosting something locally automatically removes it from a restriction defined by technology provenance. A change in hosting can change the data flow; it does not, by itself, change the model’s origin or the wording of an order.
The stated concern is about data and control
DHS cites risks associated with DeepSeek-hosted services, sensitive data and potential access under China’s legal framework. That is the department’s stated rationale, rather than proof that every installation has transferred government information.
In its October 20 response, China’s foreign ministry opposed what it described as an overstretched national-security argument and the politicisation of technology and trade.
For me, the useful distinction is between a government’s risk decision and a technical finding about a particular deployment. Both can matter, but one should not be presented as the other.
The supplier questions I take from this
Even outside federal procurement, this is a good reason to ask a little more before connecting a new AI service to work:
- Which model providers are used, and can the service switch between them?
- Where are prompts, files and outputs processed and retained?
- Which subcontractors or connected tools receive the information?
- Can an administrator restrict providers and receive notice when the arrangement changes?
- Can the organisation export its work and leave without losing essential records?
For someone working within the covered federal environment, the actual order and the organisation’s procurement and security teams should determine compliance. A consumer chatbot comparison cannot resolve that question.
Price and capability are only part of choosing a tool
I recently looked at DeepSeek’s task costs. This is a different part of the same decision: an attractive price does not answer whether a service fits the rules and data requirements of the work.
I want clear information about the supplier chain before a tool becomes difficult to replace. That is useful preparation regardless of which company provides the model.
.png)
Comments
Post a Comment