On October 20, the U.S. Department of Homeland Security (DHS) published an order signed by Secretary of Homeland Security Markwayne Mullin, requiring federal civilian agencies to remove technology associated with Chinese AI company DeepSeek from their IT systems.

Citing "supply chain security risks," the order also bars these agencies from procuring IT products and services developed by DeepSeek and its affiliates.

The move marks a further escalation in the U.S. government's restrictions on Chinese AI companies. According to foreign media reports, the U.S. Navy, Department of Defense, Department of Commerce and several state governments had previously restricted the use of DeepSeek's large language models on their devices and networks. The latest order extends the restrictions to nearly all federal civilian agencies, beyond the defense and intelligence sectors.

Of particular concern is the inclusion of "third-party products and services" built using DeepSeek technology. This means that software and other products developed by U.S. companies using DeepSeek's models could also be excluded from government procurement. The measure goes well beyond what might be justified by so-called security concerns. It effectively uses access to the substantial government procurement market as leverage to pressure U.S. companies into reducing their use of DeepSeek-related products and services.

Using such measures to suppress foreign technology companies is a familiar tactic of the U.S. government, and it often foreshadows broader restrictions and tougher sanctions. In recent years, U.S. actions against Chinese telecommunications companies Huawei and ZTE, as well as Russian cybersecurity company Kaspersky, have followed a similar trajectory, beginning with restrictions on government procurement before escalating to broader market exclusion and export controls.

The U.S. move is therefore highly likely to create a chilling effect. U.S. companies will inevitably worry about whether DeepSeek will face further restrictions and whether similar limits will extend to other Chinese AI companies, potentially leading them to reduce the position and share of Chinese AI technology in their supply chains. This may be precisely one of the deeper objectives the U.S. government seeks to achieve.

As for the security concerns regarding DeepSeek mentioned in the DHS document, these are entirely deliberate smears without any evidence. For example, the document emphasizes that DeepSeek's large language models will transmit sensitive data to China. Yet it is well known that DeepSeek's open-source models can be "locally deployed," allowing them to run entirely on an enterprise's or government agency's internal network without transmitting data externally.

On October 13, MIT Technology Review reported that if businesses want to use AI to process sensitive data or proprietary information, running a local model on their own hardware may be the best way to keep their operations secure. As for the allegation that DeepSeek models could contain malware, this is even more baseless. The series of large language models released by DeepSeek have been downloaded, used, tested, and studied by a large number of developers, and their security has long been widely verified and recognized by the global open-source community.

Security rhetoric cannot obscure the broader implications of Washington's efforts to constrain Chinese AI companies. The more the U.S. government erects barriers to the use of Chinese models, the more important it becomes for Chinese AI companies to remain committed to open innovation and contribute to the global open-source ecosystem, offering the world more capable and accessible technological alternatives.

DeepSeek offered a timely response on the very day the DHS published the order: It open-sourced DeepInfer, its internal inference framework, designed to improve the efficiency of model inference in local deployments and reduce computing and energy costs.

The U.S. may be able to prevent some government agencies from using Chinese large language models, but it cannot block the global demand for open-source AI technology that can empower economic development and make people's lives more convenient. Nor can it stop Chinese AI companies from continuing to move forward and reach the world.