
Federal civilian agencies now face a clear procurement deadline to remove DeepSeek products, stop purchasing them and examine third-party services built with the models. The Department of Homeland Security published the order in the Federal Register on October 20, 2026, citing risks to government data and infrastructure.
What does the DeepSeek order cover?
The notice applies to information technology systems across the federal civilian executive branch. Agencies must remove products affiliated with DeepSeek and are barred from purchasing the company’s products or services.
The scope extends beyond DeepSeek-branded tools. Third-party services developed using DeepSeek technology can fall within the restriction, so federal contractors may need to examine services supplied to civilian agencies.
The official notice is listed as document 2026-21447, docket DHS-2026-0628 and citation 91 FR 66346. It runs five pages, from pages 66346 through 66350.
Why did DHS act?
DHS said DeepSeek’s models present several risks to federal systems. According to the source coverage, the agency warned that the technology could track or digitally fingerprint users and could transfer sensitive data to the People’s Republic of China when used by a U.S. government agency or vendor.
DHS based the order on the 2018 Federal Acquisition Supply Chain Security Act after a review by the Federal Acquisition Security Council. The law gives the government a formal basis for excluding products and services from federal procurement.
The order does not apply to private companies, individual consumers or every state government. Its direct effect is limited to federal civilian agencies, but the contractor clause could affect private businesses that sell software or services built on DeepSeek models to those agencies.
Why did DeepSeek gain attention?
DeepSeek became widely known after releasing its R1 reasoning model in 2025. The source article says the open-weight system was trained using less powerful chips and at a lower cost than some competing models, while outperforming models from OpenAI and Anthropic in several complex tests.
A 2026 review in Frontiers in Artificial Intelligence found that DeepSeek had become the most downloaded app across 156 countries. The review also reported that early academic writing on DeepSeek was generally more positive than early writing on ChatGPT, although political and ethical questions appeared less often in the DeepSeek research.
The same review said DeepSeek’s downloadable weights allow users to run the model locally, which can give organizations more control over data. The authors also noted a limitation: DeepSeek is not fully open source under the working definition used by the Open Source Initiative because its training data have not been made entirely available.
What security concerns have been raised?
A 2025 Cornell Law School review described DeepSeek as offering capabilities comparable to major U.S. AI systems at a fraction of the development cost. It reported that DeepSeek surpassed ChatGPT in downloads within a week of its January 2025 app launch.
The review also cited a Cisco study using algorithmic jailbreaking techniques. In that research, DeepSeek failed to block any of 50 harmful prompts, while ChatGPT blocked 86 percent. The Cornell review treated the result as a concern about safety controls, not as proof that every deployment of DeepSeek would behave in the same way.
The review also raised questions about censorship and data handling. It reported that DeepSeek’s privacy policy stated that personal information was stored on secure servers in China. Local deployment can reduce some data-leakage risk, according to the review, but the model still needs careful evaluation before use with sensitive information.
What earlier restrictions preceded the order?
The source article says the Commerce Department banned DeepSeek from government devices in 2025. The U.S. Navy also directed members not to use the model because of potential security concerns. New York State banned the app from government devices and networks, citing data privacy and censorship risks.
The federal order expands those earlier measures by using a government-wide procurement rule. That matters for contractors because a product may not need to be a DeepSeek-branded application to be included if it was developed using the company’s technology.
What to do now
- Inventory any AI services used by your organization or supplied to a federal civilian agency.
- Check whether those services were developed with DeepSeek models, even if the final product uses a different brand.
- Review contracts and procurement records for tools that may fall within the order’s scope.
- Prepare replacement options before a federal customer asks for a compliant service.
- Document how personal, customer and government information is handled when a service uses an AI model.
FAQ
What did the Trump administration order regarding DeepSeek?
The Department of Homeland Security ordered federal civilian agencies to remove DeepSeek-related products from their information technology systems and stop purchasing DeepSeek products and services.
Why is the U.S. government concerned about DeepSeek?
DHS cited risks that the technology could track or digitally fingerprint users and could transfer sensitive data to the People’s Republic of China when used by a U.S. government agency or vendor.
Has DeepSeek been restricted in the U.S. before?
Yes. The Commerce Department banned DeepSeek from government devices, the U.S. Navy directed members not to use the model, and New York State banned it from government devices and networks over data privacy and censorship concerns.
Related coverage
Sources
See our editorial disclaimer for how our articles are produced.
Run a free scan to see your AI Visibility Score, SEO rating, and local citation accuracy.