DeepSeek’s rise in early 2025 — driven by low-cost, open-weight models that challenged major U.S. rivals — put the Chinese startup at the center of an already fierce technology contest.¹

By July, the Department of Homeland Security had signed an order directing civilian agencies to remove DeepSeek-related technology. The concern was not confined to the company’s chatbot: agencies were also told to identify third-party products and services built using DeepSeek models, while exceptions could be requested.²

The administration’s case is rooted in data access. DHS warned that DeepSeek’s China-based operations could allow sensitive information submitted by U.S. government users to become accessible to Chinese authorities under Chinese law. The later restrictions bar agencies from installing, accessing or running DeepSeek applications, and from awarding contracts that involve them.³

Washington’s move therefore treats the model itself — not merely a branded app — as a potential government-network risk. That is a significant escalation in the AI competition: open models can travel through other companies’ software, making a conventional product ban far easier to evade.

Beijing rejects the premise. Foreign Ministry spokesperson Mao Ning said China opposes “the overstretching of the concept of national security and the politicizing of science, technology, economic and trade issues,” and said it would protect Chinese companies’ legitimate interests.⁴

The ban also fits a broader White House argument that restraint can carry strategic costs. President Donald Trump has played down tighter checks on AI development, warning against surrendering America’s edge to China; David Sacks amplified a similar view, saying Chinese-model dominance would carry its own risks.David Sacks (DavidSacks)⁵

For now, both sides frame the dispute as defensive. The U.S. says it is shielding federal data; China says it is resisting political discrimination. Neither framing leaves much room for technological coexistence.