WASHINGTON — The Department of Homeland Security has ordered federal civilian agencies to purge all artificial intelligence tools developed by the Chinese firm DeepSeek from their information technology networks, according to a government notice made public on Monday.
The sweeping directive applies across civilian entities such as the Department of State, the Environmental Protection Agency, the Federal Communications Commission, the Department of the Treasury, and the Department of Justice, BBC reported. Agencies under the Department of Defense are excluded from the mandate.
Data Privacy and Malware Fears
Under the order published in the Federal Register, civilian departments must eliminate any existing DeepSeek applications, cease all current use, and bar third-party services that depend on the company’s AI models. The regulation also blocks DeepSeek from competing for federal government contracts, according to Engadget.
Homeland Security officials warned that operating the Chinese software poses severe cybersecurity hazards. DHS noted that the tools could potentially lead to the transfer of sensitive government data and contractor records to Beijing, while warning that the software could harbor malware or underlying vulnerabilities.
In a memorandum dated July that accompanied the directive and was made public Monday, Homeland Security Secretary Markwayne Mullin outlined specific tracking concerns, Gizmodo reported.
“DeepSeek’s account generation process collects personally identifiable information (PII) and may also track and digitally fingerprint users,” Mullin stated.
Although the public filing emerged on Monday, the operational ban took effect at the end of July, according to BBC. Federal agencies like the Navy and the Department of Commerce had already previously cautioned staff against using the Chinese startup’s software.
Rivalry Over Cheaper AI Models
DeepSeek emerged as a prominent rival to leading American tech firms after releasing high-performing models developed at a fraction of the cost incurred by U.S. competitors. When DeepSeek introduced its R1 model early last year, the company disclosed development expenses of roughly $6 million, compared to an estimated $100 million spent to train OpenAI’s GPT-4, sparking sharp market sell-offs across the U.S. tech industry, BBC reported.
The DHS action follows the release of DeepSeek’s newer V4.1-Flash-1017 model last week, which the company promoted as faster and more capable. Despite the restrictions, DeepSeek models remain widely used within the private sector, appearing on consumer app charts and powering commercial offerings by companies including Snowflake, Vercel, and OpenRouter, according to BBC.
However, the rapid growth of Chinese AI tools has triggered friction in Washington. American technology companies such as Anthropic and OpenAI have accused competitors in China of illicitly using their model outputs to train rival software, a practice known within the industry as distillation, Gizmodo reported. Last month, the FBI, the Cybersecurity and Infrastructure Security Agency, and the National Security Agency released a joint statement accusing Chinese companies of carrying out widespread distillation efforts, prompting Treasury Secretary Scott Bessent to threaten punitive sanctions.
White House tech lead Michael Kratsios previously cautioned in an internal communication that foreign entities were engaged in the exploitation of American artificial intelligence development, according to BBC.
Export Controls and Looming Oversight
Efforts to regulate the software have also surfaced on Capitol Hill, where lawmakers introduced legislation last year seeking to block DeepSeek products from all government equipment, Engadget reported.
Tensions surrounding the software extend beyond Washington. Beijing authorities have also weighed implementing export controls to limit the international deployment of certain domestic AI models, according to Gizmodo. DHS, DeepSeek, and the Chinese Embassy in Washington did not immediately comment on the publication of the directive.
