DHS Bans DeepSeek from Federal Systems, Extends Purge to Third-Party Tools

DHS Issues Sweeping Ban on DeepSeek Across Federal Agencies

On July 24, 2026, U.S. Department of Homeland Security Secretary Markwayne Mullin signed Directive 2026-21447, which was publicly released on October 19 and took effect on October 20. The order mandates two primary actions:

Remove DeepSeek and its affiliated entities from all federal procurement channels.

Require all federal civilian agencies to purge DeepSeek products from their information systems, along with any third-party products or services developed using DeepSeek technology.

This extends the ban beyond DeepSeek's own offerings to the entire ecosystem of tools that incorporate its models, meaning a U.S. vendor cannot sell a product to the federal government if its underlying technology relies on DeepSeek.

Stated Rationale: Multiple Threat Vectors

The order cites several risk categories, summarized in the directive's risk assessment:

Risk Summary: Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co., Ltd. (DeepSeek) is a private company located in the People's Republic of China (PRC) that widely provides multiple AI applications for free. DeepSeek's models present multiple threat vectors. DeepSeek's servers are located in China, and China's legal framework grants government cyber and intelligence agencies broad authority to access private company infrastructure. For DeepSeek-hosted model instances, the account generation process collects personally identifiable information (PII) and may enable user tracking and digital fingerprinting. DeepSeek's most widely used applications operate via a prompt-and-response process that requires submitting information to the model; if used by U.S. government employees or contractors, this could result in sensitive data transmission to China. Such transmission, whether alone or combined with public data, could generate actionable intelligence targeting U.S. personnel and/or government programs. The models themselves may contain malware or other security vulnerabilities. In fact, open-source security research has discovered multiple critical vulnerabilities in DeepSeek's iOS mobile application.

The directive also notes that the 60-day exemption window closed before the order was made public, effectively preventing agencies from seeking waivers.

Broader Context: Escalating U.S. Restrictions on Chinese Tech

This move follows a pattern of U.S. actions against Chinese technology firms: Huawei faced chip supply cuts, TikTok was forced into a U.S.-controlled joint venture, and now DeepSeek is targeted. Prior to this federal directive, the U.S. Department of Commerce, the Navy, and several states including New York, Tennessee, and Virginia had already restricted DeepSeek use on government devices. The DHS order consolidates these scattered bans into a unified federal mandate.

However, the article notes that DeepSeek's R1 model achieved performance comparable to OpenAI's o1 at a significantly lower cost, shaking Silicon Valley. The author questions whether administrative orders can ultimately block market adoption when a product offers superior cost-performance.