"DeepSeek’s models present 'multiple threat vectors' to civilian federal agency IT systems," Secretary of Homeland Security Markwayne Mullin wrote, ordering the exclusion of DeepSeek products on June 26 and setting a ban that federal civilian agencies must now implement.
Secretary of Homeland Security Markwayne Mullin's June 26 exclusion order
The Department of Homeland Security has issued an order, included in a notice set to publish Tuesday in the Federal Register, that bars civilian federal agencies from buying information technology products and services from Hangzhou DeepSeek Artificial Intelligence, the parent company of the AI model DeepSeek. The notice cites the Federal Acquisition Supply Chain Act and a review by the Federal Acquisition Security Council as the legal basis for the prohibition. The record includes Mullin’s June 26 letter directing the exclusion, which indicates the ban has been in place for months.
Why DeepSeek was singled out
Mullin’s letter lays out specific concerns. He wrote that some threat vectors were "unique to the company’s country of origin," while others raised potential privacy and mass data-collection issues that could apply more broadly to other Chinese and U.S. AI models. According to the order, DeepSeek models store data on servers based in China, where national security laws give the Chinese government significant latitude to request and obtain data from domestic companies.
The notice states models hosted directly by DeepSeek collect personally identifiable information from users and can trace their "digital fingerprints" online. It also notes that DeepSeek relies on a "prompt and response" process — a characteristic the secretary said "may result in the transfer of sensitive data" to the Chinese government. Mullin wrote that the transfer of such information "could produce actionable intelligence on U.S. persons and/or government programs."
The cyber insurance questionnaire just landed. Now what?
SOC 2, HIPAA, insurance renewals - someone has to own security strategy. Nubivance provides fractional CISO leadership without the full-time salary.
Get a security leadProcurement exclusion, removal obligations, and exceptions
The order requires civilian agencies to exclude DeepSeek products from future procurements and to remove any versions already deployed prior to the ban. Agencies may request exceptions, but those requests must be submitted 60 days in advance and must provide a "compelling justification," such as an inability to fulfill a mission‑critical function or other national security reasons. The notice confirms that, in previous cases, federal agencies have exercised discretion to block use of models like DeepSeek — a practice the Department of Commerce confirmed to FedScoop last year.
Operational guidance for agencies during removal
Mullin encouraged agencies that are removing DeepSeek to enhance security controls to mitigate potential compromise. The letter suggested concrete steps: monitoring the flow of traffic in and out of agency networks, implementing enhanced logging and controls on data exfiltration for relevant systems, and detecting and flagging anomalous activity. Agencies were urged to take these measures "during the period between the issuance of this order and full removal of covered articles from agency systems."
How technologists, procurement teams, and national security officials will respond
- Technologists and security teams: They are being asked to monitor traffic, increase logging and data‑exfiltration controls, and flag anomalous behavior on affected systems while removing DeepSeek-hosted models.
- Procurement officers and acquisition officials: They must exclude DeepSeek products from future solicitations and handle removal of any deployed versions; they can pursue exceptions only with 60 days' advance notice and a "compelling justification."
- National security and policy officials: The order frames the concern as both country-specific and model‑agnostic, noting DeepSeek by name while referencing similar worries about other Chinese open‑weight AI models — including Moonshot AI (Kimi) and Alibaba (Qwen) — that have troubled national security officials or been accused of illicitly distilling U.S. models.
CyberScoop reported that the decision singles out DeepSeek — described in the order as one of the most popular and powerful Chinese open‑weight AI models — as the only Chinese model named in this action, while noting other models have drawn scrutiny. CyberScoop also reached out to CISA and DHS for further details.
The ban’s next formal step is publication in the Federal Register. For agencies, the practical effects are immediate: cease procurement, remove existing instances, document any exception requests with a 60‑day lead time and a compelling justification, and follow the security measures Mullin encouraged to reduce the risk that data already handled by DeepSeek could be exploited.
