Opens in a new tab

Trump administration orders DeepSeek banned from federal civilian agencies

  • Home
  • Blog
  • Trump administration orders DeepSeek banned from federal civilian agencies
DeepSeek ban depicted as AI whale locked inside federal vault

The Department of Homeland Security ordered federal civilian agencies to remove DeepSeek products and services from their information technology systems and stop purchasing them. For small businesses, consider using the order as a model for reviewing AI tools, embedded third-party services, and the company information those tools receive.

What did the federal order require?

The Department of Homeland Security published the order on October 20, 2026, under Docket No. DHS-2026-0628. The notice runs five pages, from pages 66346 to 66350, and carries document number 2026-21447 and citation 91 FR 66346. According to the Federal Register document, civilian agencies must remove DeepSeek products from their IT systems, prohibit future purchases of DeepSeek products and services, and exclude DeepSeek from federal contracts.

The order is broader than removing the DeepSeek mobile application. It reaches affiliated products and third-party services built with DeepSeek models. Review the underlying technology rather than relying on the visible product name.

Why was the order issued?

DHS stated that DeepSeek’s models present “multiple threat vectors.” The agency cited the 2018 Federal Acquisition Supply Chain Security Act and a review by the Federal Acquisition Security Council.

The notice also raised several security concerns. DHS said DeepSeek models could be used to “track or digitally fingerprint” users and warned about the possible “transfer of sensitive data to the People’s Republic of China if used by U.S. government or vendors.” Homeland Security Today reported that DHS also said the models could contain malware or other security vulnerabilities.

Those points are DHS’s stated findings and risk assessment. They are separate from the order’s operative requirements, which concern removal, procurement and contract participation. Consider them as DHS’s stated rationale rather than proof about a particular service.

Why supplier transparency matters

According to SiliconANGLE, the federal order covers products affiliated with DeepSeek, including third-party services developed using the company’s models. Cornell Law School’s 2025 analysis, U.S. AI Policy and the DeepSeek Problem, reported that DeepSeek’s privacy policy states that personal information is “stored in secure servers located in the People’s Republic of China.” The analysis also discussed control of the application by the People’s Republic of China and a lack of safeguards against malicious modification as risks for consumers and national security.

That same Cornell article cited a Cisco evaluation using algorithmic jailbreaking. Consider the result in the context of the cited test rather than treating it as a finding about every prompt or deployment.

Self-hosting can reduce exposure to data leakage, according to the Cornell analysis. A 2026 Frontiers in Artificial Intelligence mini review likewise described self-hosting as a way for organizations to control data governance and customization. The review also cautioned that DeepSeek is not fully open source because its training data have not been made entirely available.

Context, not findings

DeepSeek was founded in 2023 and is based in Hangzhou, China, according to the Congressional Research Service. Its V3 model has 671 billion parameters, and DeepSeek reported benchmark scores comparable to OpenAI’s GPT-4o and Anthropic’s Claude 3.5 Sonnet. The Congressional Research Service described R1’s evaluations as covering general reasoning, graduate-level questions and mathematics problem solving.

These technical details provide context, not a conclusion about a particular private deployment. For a small business, the more useful fact remains narrower: the federal order expressly includes third-party services developed with DeepSeek models.

Small-business AI supplier review

Use these fields to structure your own review.

  • Service and vendor: Record the product, provider, business owner and how employees access it. Include browser tools, mobile and desktop applications, coding assistants, document-processing services, embedded chatbots and services supplied by an outside vendor.
  • Model or API disclosures: Ask the vendor which models and APIs power the service, including models used by subprocessors or added in later updates. Confirm where submitted information is processed and retained. Review the underlying technology rather than relying on the visible product name.
  • Information types submitted: List customer records, employee details, contracts, financial information, credentials and source code that may enter the service. Review administrator controls, account access and logging for each approved use.
  • Approved owner and review date: Name the person accountable for approval and record the next review date. If the vendor cannot explain what technology processes company information or where the information goes, hold approval for sensitive work until those questions are answered.

What to do now

  1. Create an AI inventory. Ask employees and office managers which AI applications, browser services, coding assistants and vendor tools they use for company information. Include tools embedded in software that staff may not recognize as AI products.
  2. Request model and data details. Ask each supplier which models and APIs it uses, where information is processed, what is retained and whether a subprocess or model can change later. Record the answers instead of relying on an application name.
  3. Match approval to information sensitivity. Review whether staff can submit customer records, employee details, contracts, credentials or source code. Restrict sensitive uses where the supplier cannot provide a clear account of data handling and model use.
  4. Assign ownership and a review date. Give every approved service a named administrator or business owner. Revisit the decision when the product, its underlying model, contract terms or data-processing arrangements change.
  5. Keep local deployment in the same review. According to the Cornell analysis, running DeepSeek offline reduces the risk of data leakage. Consider access controls, patching and an approved list of information that staff may submit as part of the review.

FAQ

Does the DeepSeek ban apply to private businesses?

No. The October 20, 2026 Federal Register notice applies to federal civilian agencies, their IT systems and federal contracts. It does not directly prohibit private businesses from using DeepSeek.

Does the order cover third-party services that use DeepSeek models?

Yes. The order covers products affiliated with DeepSeek, including third-party services developed using the company’s models. Review the underlying technology rather than relying on the visible product name.

Should a small business remove DeepSeek products?

The federal order does not require that removal from private businesses. The practical response is to inventory AI services, request model and data-processing disclosures, identify the information submitted to each service, and assign an owner and review date for every approval.

BizScoreAI

BizScoreAI, which includes the free listing check

BizScoreAI scores how visible a business is to AI search and shows what its listing looks like to the engines people ask. Open the free listing check.

This article reports on publicly available information from the sources listed below. We have no reason to believe it is inaccurate, but we have not independently verified every claim, and details may change as the story develops.

Sources

← All Articles