US Orders 60-Day DeepSeek Purge From Federal Agencies as DHS Cites China Security Risks
Updated
Updated · InfoWorld · Oct 21
US Orders 60-Day DeepSeek Purge From Federal Agencies as DHS Cites China Security Risks
3 articles · Updated · InfoWorld · Oct 21
Summary
Federal civilian agencies must remove all DeepSeek products and services from their IT systems within 60 days under a Homeland Security order, with only written CISA exemptions allowed for compelling mission needs.
DHS said China-hosted DeepSeek creates multiple threat vectors: possible transfer of sensitive prompts and personal data to the PRC, broad Chinese government access powers, user fingerprinting, and potential malware or other vulnerabilities in the models.
The ban also covers affiliated products and third-party tools built on DeepSeek, widening the impact to software supply chains across agencies including Treasury, State and Justice, while excluding DoD national security systems and the general public.
DeepSeek faces added scrutiny because it leaked more than 1 million log entries in January 2025, even as it now open-sources its DeepInfer inference framework and promotes lower-cost enterprise AI deployment.
The move underscores a broader US push to tighten AI procurement and third-party risk controls as agencies and companies weigh DeepSeek's performance gains against data-sovereignty and cybersecurity concerns.
How will government agencies uncover foreign AI code secretly buried deep within their everyday third-party software applications?
Could banning highly advanced foreign open-weight models inadvertently trap federal agencies in a costly technological disadvantage?
What hidden technical signatures reveal that a seemingly harmless AI tool is actively routing sensitive data overseas?
U.S. Bans DeepSeek AI: Inside the 2026 DHS Order, Compliance Fallout, and Global Tech Realignment
Overview
On October 19, 2026, the Department of Homeland Security issued a sweeping order requiring all federal agencies to remove DeepSeek’s Chinese-developed AI technology, following a formal recommendation from the Federal Acquisition Security Council. This action was driven by concerns that Chinese laws allow government access to corporate data, creating risks of sensitive U.S. information being transferred abroad. The ban also responded to evidence of DeepSeek’s covert extraction of proprietary reasoning from American AI models. As a result, federal IT teams began auditing software for hidden dependencies and migrated to approved alternatives, while the order set a benchmark that influenced state governments and global investments in sovereign AI infrastructure.