US Bans Use of China’s DeepSeek Across Civilian Agencies
The US has ordered civilian government agencies to remove DeepSeek AI tools and technology built with its models from their systems. The Department of Homeland…
The US bans use China’s DeepSeek tools across civilian government agencies, with the Department of Homeland Security warning that using them may expose sensitive data to China. The order also requires agencies to remove technology built with DeepSeek models from their systems.
The restriction affects civilian agencies, including the Departments of the Treasury, State and Justice, but excludes agencies within the Department of Defense. The BBC reported that the ban took effect at the end of July, before the DHS made the order public on Monday.
Data security at the center of the order
A document attached to the DHS notice said DeepSeek’s models “present multiple threat vectors.” The department warned that use by government employees or vendors “may result in the transfer of sensitive data to [China]”.
The Guardian reported that the notice also pointed to DeepSeek’s servers being located in China. The department said China’s legal system gives its government “extensive authority” to access private company infrastructure. DeepSeek had been contacted for comment.
The measure reaches beyond DeepSeek’s own products. It also covers tools developed by third parties using the company’s AI, meaning agencies must consider not only which chatbot or model they access but whether DeepSeek technology sits within another service.
That distinction matters for public-sector technology use. A product may not carry DeepSeek’s name, yet still fall under the order if its development relied on the company’s models. The directive puts responsibility on agencies to identify and remove both kinds of tools from their IT devices.
Wider effects for AI developers
DeepSeek has attracted users with open-source models that are free to access and less expensive to use, according to the BBC. The company’s technology has also been incorporated into offerings from US technology firms, including Snowflake, Vercel and OpenRouter.
The ban is limited to civilian government agencies, so it does not itself prohibit private companies from using DeepSeek. Still, the DHS action may affect vendors that work with federal agencies if their products include technology developed with the models. The notice’s scope makes the origin of a tool relevant, even when an agency accesses it through a third party.
There is a cost dimension to the dispute. The BBC reported that DeepSeek models can cost developers a fraction of closed alternatives, and that startups have flocked to open-source models. A government restriction on their use could therefore shape which AI tools agencies and their suppliers can deploy, though the sources do not specify how agencies will replace them.
DeepSeek’s rise has also become part of a broader US-China contest over artificial intelligence. The Guardian reported that US officials have accused DeepSeek and other leading Chinese AI firms of systematically stealing capabilities developed by American rivals. Those are US allegations, not findings detailed in the DHS notice.
According to the BBC, Anthropic and OpenAI have also claimed that Chinese AI firms, including DeepSeek, copied their models. The administration has separately criticized Chinese AI firms, while US President Donald Trump has described China as the country’s key rival in AI development.
The order applies to civilian agencies, not the Department of Defense. The DHS warning is that DeepSeek use may transfer sensitive government data to China.
Source: theguardian.com