US Orders DeepSeek AI Removed From Civilian Government Systems

The United States has ordered civilian government agencies to remove artificial intelligence (AI) technology from Chinese company DeepSeek, citing ongoing concerns surrounding Chinese AI companies.

The Department of Homeland Security (DHS) has banned the use of DeepSeek across civilian government agencies, as well as any technology built using its AI models.

DeepSeek ban applies to civilian agencies

The order was announced by DHS on Monday, although the ban took effect at the end of July. It applies to civilian agencies including the Treasury Department, State Department and Justice Department.

Agencies within the Department of Defense are excluded from the order.

The DHS notice did not detail the specific safety concerns surrounding DeepSeek. However, the White House has criticized Chinese AI companies this year.

DHS, DeepSeek and the Chinese Embassy in Washington, D.C., have been contacted for comment.

Why is the US concerned about DeepSeek?

The ban follows criticism of Chinese AI from the Trump administration and US technology companies.

White House technology chief Michael Kratsios wrote in an internal memo earlier this year that “malicious exploitation” of US AI companies was being carried out primarily by Chinese rivals.

Anthropic and OpenAI, both planning to list on the stock market in the coming months, claim that some Chinese AI companies, including DeepSeek, have essentially copied their AI models.

US President Donald Trump has made clear that he views China as America’s main rival in AI development, which he considers critical to the country’s future.

DeepSeek’s popularity and low-cost AI models

DeepSeek has grown in popularity since it was first published in 2025. Demand for its models has been so high that the company’s servers sometimes go down.

DeepSeek’s free chatbot app is among the most downloaded apps in the Apple App Store rankings. Its AI models are also used by US technology companies, including Snowflake, Vercel and OpenRouter, which have incorporated DeepSeek into services offered to customers.

Startups are increasingly turning to open-source models such as DeepSeek.

For technology companies and AI developers, using DeepSeek for work costs a fraction of using closed models such as Google’s Gemini, Anthropic’s Claude and OpenAI’s Codex.

The cost of developing DeepSeek’s model has also been significantly lower than the amounts US AI companies have spent developing their own systems.

Last year, DeepSeek said it had spent around $6 million (£4.5 million) training the debut version of its model. By comparison, costs associated with OpenAI’s GPT-4 were around $100 million. The revelation sent technology stocks plummeting.

DeepSeek’s latest AI model

DeepSeek released its latest version last week. Users have praised the model online, saying it outperforms even OpenAI’s most advanced AI tools on common tasks such as writing computer software.