The U.S. Department of Homeland Security (DHS) published on Monday, October 19, a landmark order prohibiting federal civilian agencies from purchasing or using products developed by the Chinese artificial intelligence company DeepSeek, as well as any third-party tools built on its implementations. The order, originally signed on July 24, 2026, and made public only after nearly three months, was filed in the Federal Register under docket number DHS-2026-0628.
What makes this order different from earlier restrictions on DeepSeek is its breadth. Since early 2025, various U.S. government agencies had already banned the use of DeepSeek's chatbot and models on their devices — the Navy, NASA, the Commerce Department, and more recently the Defense Information Systems Agency, which blocked access to the models in January 2025. But those bans were limited to specific agencies or particular devices. The July 2026 order, issued under the Federal Acquisition Supply Chain Security Act (FASCSA) of 2018, covers all U.S. federal civilian agencies.
The agency described DeepSeek's models as presenting multiple threat vectors for government systems. The DHS's core justification is that, when users create accounts on the company's products, they provide personally identifiable information that can be tracked or digitally fingerprinted and later transferred to China, where the Chinese Communist Party could access it. The document states that the transfer of the aforementioned information to the PRC, either independently or in combination with publicly available data, could produce actionable intelligence on U.S. persons and government programs.
Furthermore, the DHS warns that the models themselves may contain malware or security vulnerabilities. The department cited research identifying critical vulnerabilities in DeepSeek's iOS application, whose backend, according to earlier reports, connects to China Mobile — a Chinese military company classified by the U.S. government. In 2019, the Federal Communications Commission (FCC) had already banned China Mobile from operating in the United States, warning that unauthorized access to customer data could cause irreparable damage to U.S. national security.
The most impactful aspect of the order for the technology industry is the clause covering third-party products. Agencies must not only remove direct DeepSeek applications, such as the chatbot and the coding environment, but also software from other vendors that was developed using DeepSeek implementations. The order does not precisely define what developed using implementations means, raising questions among contractors. A vendor that fine-tuned an open-weight DeepSeek model, or that simply routes some queries to its servers, could fall within the scope of the ban. This poses a direct question to any company selling AI features to government agencies: which model powers your product?
To handle this complexity, the order allows for exceptions. Agencies that believe they need a DeepSeek-based product can request a waiver from the DHS by submitting a written justification within 60 days of the order's issuance. The government also urged all affected agencies to monitor network traffic entering and leaving their systems, strengthen auditing, and protect against data exfiltration until the removal is complete.
The Chinese response was immediate. Foreign Ministry spokesperson Mao Ning said Beijing consistently opposes the overextension of the concept of national security and the politicization of scientific, technological, economic and trade issues. China will resolutely safeguard the legitimate rights and interests of Chinese companies, she said at a press briefing.
The order sits within a broader geopolitical context. DeepSeek gained global prominence in January 2025 when its R1 reasoning model demonstrated competitive performance against American models while using a fraction of the computing power. This triggered a price war across the AI sector and caused a sharp decline in Nvidia's stock. Since then, the U.S. government has intensified pressure on the company: in September 2026, the United States accused DeepSeek and five other Chinese companies of using large-scale distillation techniques to extract capabilities from American frontier AI systems. In October 2026, Beijing announced plans to restrict the export of 23 AI-related technologies, including chip designs and certain advanced models.
At a time when Presidents Trump and Xi are set to meet in mid-November at the APEC summit in Shenzhen, artificial intelligence will undoubtedly be on the agenda. The DHS order may be one of several flashpoints between the two countries, which have made progress on some cooperation fronts, such as the U.S. proposal to establish a national security incident notification system, but remain deeply divided on technology access and governance.
Sources: TechCrunch, Federal Register, CNBC
✓ Independent sources cross-checked and verified before publishing
