The US Department of Homeland Security issued an order on Monday directing federal civilian agencies to stop using technology products from Chinese AI lab DeepSeek and to remove any existing implementations.

The order, posted for public inspection in the Federal Register, covers "all DeepSeek and DeepSeek-affiliate products, and third-party products and services developed using DeepSeek implementations."

The first part of the ban targets DeepSeek's own apps and API. The second part reaches further into the AI supply chain, and the order does not define what "developed using" means. It could refer to products that call DeepSeek models through an API, or to any software whose developers used DeepSeek while writing or testing code.

AI gateway services add another complication. OpenRouter and Vercel AI Gateway both make DeepSeek's models available through their APIs, and both route requests to the provider the service considers the best fit by default. Agencies that rely on these gateways without restricting providers could end up using DeepSeek models without knowing it.

Katie Moussouris, founder and CEO of Luta Security, said banning DeepSeek's own app and API is workable, but the third-party clause is not. According to Moussouris, an AI software bill of materials does not record which tools developers used, so every vendor would have to attest that nothing it relies on ever touched DeepSeek, with no practical way to verify it.

Moussouris also said applying the order to US-hosted versions of DeepSeek models goes too far, since the risks the order cites are mostly about data flowing to DeepSeek's servers in China.

The order says only that the model may contain malware or vulnerabilities. NIST evaluations found DeepSeek more vulnerable to prompt injection than US models and easier to jailbreak. A DHS spokesperson did not respond to a request for comment.

US officials have discussed a DeepSeek ban since the lab released DeepSeek-R1 in early 2025. On February 6, 2025, Representatives Darin LaHood and Josh Gottheimer introduced the No DeepSeek on Government Devices Act, which never left committee.

The Defense Information Systems Agency started blocking DeepSeek on Pentagon networks in January 2025, and the US Navy, NASA, and the House of Representatives took similar steps, as did some state governments.

Security researchers also flagged problems in DeepSeek's mobile apps. A December 2025 report on the Android and iOS apps cited plaintext information leaks, links to servers run by Chinese tech firms, and a lack of SSL pinning.

Organizations that work with federal agencies, or that simply want to avoid DeepSeek models, can take these steps:

1. Check the provider settings in AI gateways such as OpenRouter or Vercel AI Gateway and restrict routing to approved model providers.
2. List internal tools and third-party services that call DeepSeek models directly or through a gateway.
3. Ask vendors whether their products use DeepSeek implementations, keeping in mind that the order does not define the term.

DHS has not clarified how agencies should interpret the third-party clause or whether US-hosted DeepSeek models fall under the same removal requirement.