US orders federal agencies to purge DeepSeek AI from government systems

The Department of Homeland Security has ordered federal civilian agencies to stop buying products and services from Chinese AI startup DeepSeek and to strip them from existing government IT systems, citing national security concerns tied to the company's operations in China.

The directive, announced by DHS, reaches beyond DeepSeek's own products. It also covers third-party products and services built using DeepSeek technology. Agencies can request exceptions.

The order raises the stakes in the US-China AI race, where DeepSeek's low-cost open-weight models compete directly with American rivals such as OpenAI and Meta. Open-weight means the trained model files are publicly available for anyone to download and run themselves.

What does the order actually change?

The DHS warned that DeepSeek's handling of user data could expose sensitive US government information. Under Chinese law, that data could be accessible to Chinese authorities, the order said.

This is not the first restriction. A law passed in December 2025 already targeted DeepSeek's use in intelligence community national security systems. The new order extends the ban to federal civilian agencies, covering both new purchases and existing IT systems.

In plain words: the ban now covers nearly the whole federal civilian apparatus, not just intelligence networks.

How did Beijing respond?

China pushed back quickly. Foreign Ministry spokesperson Mao Ning said Tuesday that Beijing opposes "overstretching the concept of national security and politicizing technology and trade issues," according to an official briefing.

"We will resolutely safeguard the legitimate rights and interests of Chinese enterprises," Mao added.

The rhetorical standoff arrives as both governments escalate export and procurement controls. Bloomberg reported last week that China is considering export restrictions on 23 AI-related technologies, including chip designs and certain advanced AI models, to address security and misuse risks.

Why has DeepSeek drawn global scrutiny?

DeepSeek has attracted regulatory attention since its R1 reasoning model shook the AI industry in January 2025, triggering a sharp sell-off in Nvidia shares. The episode demonstrated that a Chinese lab could match frontier capabilities at a fraction of the cost, which is precisely why Washington now treats its presence in federal systems as a data-security risk.

The US has also accused DeepSeek of intellectual extraction. According to a September report, the US accused DeepSeek and five other Chinese AI companies, including Alibaba and Moonshot AI, of large-scale model distillation — a technique that uses outputs from one AI model to train another.

American AI labs have raised the same complaint separately:

• OpenAI told US lawmakers earlier this year it had observed suspected attempts by DeepSeek to distill its models.
• Anthropic said it had identified unauthorized attempts by DeepSeek and other Chinese AI labs to collect outputs from Claude, its commercial model.

Notably, the DHS order itself does not cite the distillation allegations; it rests on data-handling concerns. Whether the ban meaningfully reduces exposure is another question — the order allows exceptions, and it does not address DeepSeek-derived open-weight models already embedded in tools agencies may not know they run.

For now, the directive signals that procurement bans, not just export controls, are becoming Washington's preferred instrument against Chinese AI — and Beijing's contemplated counter-restrictions suggest the decoupling will run in both directions.