US orders federal agencies to purge DeepSeek from government systems over security concerns

DHS orders federal agencies to drop DeepSeek

The United States has instructed federal civilian agencies to stop purchasing products and services from Chinese AI startup DeepSeek and to remove the company's technology from government IT systems, TechCrunch reported on October 20. The directive comes from the Department of Homeland Security (DHS) and rests on national security concerns tied to DeepSeek's operations in China.

The ban is wider than DeepSeek's own apps and models. According to TechCrunch, it also covers third-party products and services built using DeepSeek technology, though individual agencies can request exceptions.

Data exposure under Chinese law

The DHS's central worry is data handling. According to the order, sensitive information entered by US government users could potentially become accessible to Chinese authorities under Chinese law. That rationale, rather than any specific reported incident, is what drives the removal requirement.

An extension of earlier restrictions

The move builds on measures already in place. A law passed in December 2025 restricted DeepSeek's use in national security systems tied to intelligence work, TechCrunch notes. The new order broadens that reach to federal civilian agencies and applies both to new purchases and to systems already deployed across government IT.

Beijing pushes back

China responded sharply. Speaking at an official briefing, Foreign Ministry spokesperson Mao Ning said Beijing opposes what she characterized as overstretched national security arguments and the politicization of technology and trade issues, and said China would resolutely protect the legitimate rights and interests of Chinese enterprises, according to TechCrunch.

A wider confrontation over AI

The order lands amid several escalating disputes between the two countries. According to a September report cited by TechCrunch, 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 system to train another. OpenAI separately told US lawmakers it had observed suspected attempts by DeepSeek to distill its models, while Anthropic said it identified unauthorized attempts by DeepSeek and other Chinese AI labs to collect outputs from Claude.

China, meanwhile, is reportedly weighing its own counters. Bloomberg reported last week that Beijing is considering export restrictions on 23 AI-related technologies, including chip designs and certain advanced AI models, citing security and misuse risks.

DeepSeek has drawn government scrutiny worldwide since its R1 reasoning model upended the AI industry in January 2025, a release TechCrunch notes also triggered a sharp sell-off in Nvidia shares. The startup competes with US rivals such as OpenAI and Meta largely on the strength of low-cost, open-weight models.

Why it matters

The order shows that AI competition between the US and China is increasingly being fought through procurement rules and security policy rather than the market alone. Two details make it consequential. First, because DeepSeek's appeal rests on cheap, open-weight models, its technology can end up embedded in other vendors' products, which is precisely what the third-party clause of the order targets. That leaves agencies with a non-trivial audit task: identifying where DeepSeek-derived components sit inside tools they already run. Second, the move fits a broader pattern of escalation, from US distillation accusations and American lab complaints to reported Chinese export control plans. Whether other governments follow the US lead, and whether Beijing follows through on its own export curbs, will help determine how freely AI models and the technology behind them can cross borders at all.