DHS identified several potential security risks associated with DeepSeek's AI models, particularly the possibility that information processed through its systems could become accessible to Chinese authorities under the country's laws. The department warned that government use of the technology could expose sensitive information about U.S. citizens and federal programs.
The agency also raised cybersecurity concerns, pointing to publicly available research identifying critical vulnerabilities in DeepSeek's iOS application. According to the order, the company's models could expose government systems to malware and other security weaknesses.
Federal agencies may request exemptions or additional time to comply, provided they submit written justifications to DHS within 60 days of the directive's issuance.
The latest restrictions expand measures already taken by several parts of the U.S. government since DeepSeek attracted international attention in early 2025. The Defense Information Systems Agency blocked access to the company's models in January 2025, while Commerce Department divisions prohibited employees from using DeepSeek on government devices that March. Congress subsequently restricted the technology's use in intelligence community systems in December 2025.
Unlike those earlier restrictions, the new DHS order establishes a broader prohibition across civilian federal agencies, covering both existing technology and future purchases.
China criticized the decision, arguing that the United States was applying national security concerns too broadly to technology and trade. Foreign Ministry spokesperson Mao Ning said Beijing would defend the legitimate interests of Chinese companies.
“China consistently opposes the over-generalization of national security as well as the politicization of science, technology and trade issues,” Mao said, according to Reuters.
The directive comes as DeepSeek faces increasing scrutiny over its technology and business practices. The Chinese company gained attention in January 2025 with its R1 reasoning model, which challenged established U.S. AI developers through lower-cost, open-weight technology.
U.S. authorities have also raised separate concerns about Chinese AI companies obtaining information from American models through a technique known as model distillation. A September report accused DeepSeek and five other Chinese companies, including Alibaba and Moonshot AI, of using large-scale distillation to extract information from U.S. AI systems.
OpenAI and Anthropic have separately reported suspected unauthorized attempts to collect outputs from their models. OpenAI informed U.S. lawmakers earlier this year about suspected distillation activity involving DeepSeek, while Anthropic identified similar attempts targeting Claude.
The DHS order does not establish that DeepSeek has transferred U.S. government data to Chinese authorities. Instead, it cites potential exposure and cybersecurity vulnerabilities as grounds for removing the technology from civilian government systems.
The decision adds another federal restriction on DeepSeek as Washington and Beijing continue to compete over AI development and security. By including third-party products built with DeepSeek models, the administration is also requiring agencies to consider whether the technology is embedded in software supplied by other companies.
This analysis is based on reporting from CNBC.
Image courtesy of Jaegyeong Ilbo.
This article was generated with AI assistance and reviewed for accuracy and quality.
