Full Breakdown
U.S. Homeland Security Bans Chinese AI Model DeepSeek from Civilian Federal Agencies
By Drooid · · How we work
DHS Order Removes DeepSeek from Civilian Agency IT Systems
On October 19, the Department of Homeland Security (DHS) issued a notice in the Federal Register directing all civilian federal agencies to eliminate DeepSeek products and any third-party services built on the Chinese AI model from their information-technology environments. The order also bars future acquisition of DeepSeek-derived technology. The directive references a June 26 letter from DHS Secretary Markwayne Mullin that first ordered the exclusion, and a July 24 signing of Exclusion and Removal Order #2026-01, which formalized the procurement ban. The ban applies to agencies such as the Treasury, State and Justice departments, while Department of Defense networks remain exempt.
Background and Prior Restrictions
DeepSeek, a Beijing-based AI firm that released its open-weight models in 2025, has faced escalating U.S. scrutiny. The Commerce Department barred the model from government-furnished equipment in March 2025, and the Navy issued a device-level ban later that year. Several states—including New York, Tennessee and Virginia—have independently prohibited DeepSeek on government devices. The current DHS order expands earlier agency-level actions into a comprehensive supply-chain restriction.
Security Concerns Cited by DHS
DHS characterizes DeepSeek’s models as presenting “multiple threat vectors.” Specific concerns include:
- Data residency – DeepSeek’s servers reside in the People’s Republic of China (PRC), where national-security laws permit the government to compel private companies to hand over data.
- Personal data collection – The model’s account-generation process gathers personally identifiable information and creates “digital fingerprints” of users.
- Prompt-and-response exposure – Every user query is transmitted to the model, creating a pathway for sensitive information to be transferred abroad.
- Potential malware – Open-source security research has identified critical vulnerabilities in the DeepSeek iOS application.
These factors, DHS argues, could enable the PRC to compile actionable intelligence on U.S. persons and government programs.
Official Statements & Responses
The Chinese Embassy in Washington declined to comment when contacted. No additional federal agency has publicly challenged the ban.
Data & Statistics
- DeepSeek’s debut model cost roughly $6 million to train, compared with about $100 million for OpenAI’s GPT-4.
- The company’s V4.1 Flash model, released on September 10, is priced at $0.15 per million input tokens, undercutting most U.S. competitors.
- DeepSeek’s free chatbot ranks among the most downloaded apps in the Apple App Store and is integrated into services offered by firms such as Snowflake, Vercel and OpenRouter.
Impact on Federal Procurement
The Exclusion and Removal Order prohibits civilian agencies from contracting with DeepSeek or any affiliated entity. Agencies may request exceptions, but must submit a “compelling justification” at least 60 days in advance. The ban does not extend to the Pentagon’s separate restrictions, nor does it place DeepSeek on the Commerce Department’s Entity List, which would trigger broader export-control measures.
Conflicting Reports & Gaps
Additionally, DHS did not disclose the specific technical vulnerabilities that prompted the malware allegation, leaving a gap in publicly available detail.
Verbatim Quotes
- “The transfer of the aforementioned information to the [People’s Republic of China], either independently or in combination with publicly available data, could produce actionable intelligence on U.S. persons and/or government programs,” — Homeland Security Markwayne Mullin, homeland security secretary