Homeland Security Orders DeepSeek Purge From Federal Systems

The U.S. Department of Homeland Security has ordered federal civilian agencies to remove DeepSeek products from their IT systems, citing "multiple threat vectors" from the Chinese AI startup's models.

The order, dated to late July and made public Monday in the Federal Register, goes beyond the company's own software. It also covers third-party products and services built using DeepSeek, and it blocks civilian agencies from procuring anything tied to the firm.

China answered within a day. Foreign Ministry spokesperson Mao Ning told reporters at a routine briefing in Beijing on Tuesday that China "will resolutely safeguard the legitimate rights and interests of Chinese companies."

"China consistently opposes the overstretching of the concept of national security and the politicizing of science, technology, economic and trade issues," Mao said.

DeepSeek did not immediately respond to a request for comment.

Why does the ban reach beyond DeepSeek's own apps?

DeepSeek is one of China's best-known AI companies. Its open-source models are widely available, including in the United States, and mostly free to use. That reach is precisely what worries Washington.

The Homeland Security notice names two concrete concerns. First, DeepSeek's servers sit in China. Second, Chinese authorities could potentially access private company infrastructure and sensitive U.S. data through that setup.

The sweep of the order matters more than the target. Any vendor that built a product on top of DeepSeek's models now faces the same federal exclusion. An agency running a contractor's tool with DeepSeek inside must pull it, too.

What does this mean for U.S.-China AI relations?

The dispute lands weeks before a scheduled summit. President Donald Trump and Chinese President Xi Jinping plan to meet in mid-November on the sidelines of the APEC leaders' summit in Shenzhen, China. AI is expected to be on the agenda.

The two countries have made some recent progress on AI governance. Washington proposed a notification system for national security-related AI incidents. Both sides have also worked with the United Nations on AI cooperation tracks.

But the friction list keeps growing:

• U.S.-led restrictions still bar China from some of the world's most advanced AI chips and the machines that make them.
• Beijing last week proposed new plans to restrict exports of certain advanced AI models.
• The U.S. has accused Chinese AI startups, including DeepSeek, of large-scale "distillation" — extracting capabilities from frontier American AI models. Beijing dismissed the claims.
• The Federal Communications Commission earlier this year barred imports of new foreign-made humanoid robots, a move widely seen as targeting China.

What comes next?

The DeepSeek ban shows the mechanism Washington now favors: procurement blacklists and Federal Register notices rather than broad legislation. Every federal agency must audit its stack, and every vendor selling into government must prove its models carry no DeepSeek lineage.

The Shenzhen summit in mid-November will test whether governance cooperation survives an enforcement action aimed squarely at China's most famous AI export.

By Daniel Okafor