US bans DeepSeek from civilian agencies, but order leaves questions unanswered

Civilian federal agencies in the US may no longer use technology from the Chinese AI company DeepSeek. That is according to the tech outlet The Decoder, which relays the Wall Street Journal's coverage of the order in the Federal Register, where US authorities publish official decisions.

The ban does not apply to private companies, only to civilian federal agencies.

Affiliated companies and third-party services also covered

According to The Decoder, the ban also covers companies affiliated with DeepSeek and third-party services built on DeepSeek's technology. That means it is not only DeepSeek's own chatbot that is affected, but potentially also other services that have DeepSeek under the hood.

The order is based on a 2018 US law on supply chain security in government procurement. Rules of this kind are meant to ensure that the government does not buy products or services that could pose a risk.

Unclear how far the ban reaches

The order sets no deadline for when the technology must be removed. It is therefore impossible to tell how quickly agencies must act.

The Decoder also notes that it is unclear how the phrase "built with" DeepSeek should be interpreted. As a result, it is also unclear whether AI gateways such as OpenRouter are covered. An AI gateway is a service that provides access to multiple different AI models in a single place.

The Decoder also writes that the department has not publicly stated which security concerns were behind the decision.

What does it mean for readers?

The ban primarily affects US government employees and the systems they use. Danish users are not directly affected. The case does show, however, that the origin of AI models is increasingly becoming a matter of security and procurement rules.

The information comes from The Decoder's account of the Wall Street Journal's coverage. The order itself has not been reviewed here.