US bans Chinese AI startup DeepSeek from civilian government agencies
The Department of Homeland Security ordered federal departments to purge DeepSeek products and barred the firm from government contracting over data security concerns.
Listen
The rundown
- The Department of Homeland Security published a federal notice on Monday, Oct. 19, 2026, banning civilian government agencies from using artificial intelligence tools made by Chinese startup DeepSeek.
- The directive, which initially took effect at the end of July, requires agencies to remove DeepSeek products and third-party software built with its models from government IT systems.
- The ban bars DeepSeek products from federal contracting and covers civilian agencies including the State, Treasury, and Justice departments, but excludes military and intelligence agencies.
- The DHS stated that DeepSeek models present multiple threat vectors and could transfer sensitive U.S. government data to servers in China.
- The order comes ahead of a scheduled November 2026 summit in Shenzhen between President Donald Trump and Chinese President Xi Jinping, where AI competition is expected to be addressed.
The U.S. Department of Homeland Security published a federal notice on Monday, Oct. 19, 2026, ordering civilian government agencies to remove artificial intelligence tools developed by the Chinese startup DeepSeek from their information technology systems. The directive also prohibits agencies from using any third-party software built with DeepSeek models and bars the company's products from U.S. government contracting.
The order initially took effect across federal civilian departments at the end of July before the department made the directive public on the Federal Register website on Monday, Oct. 19. Because the mandate was issued under Homeland Security authority, it applies to civilian bodies such as the Treasury, State, and Justice departments. It excludes the Department of Defense and U.S. intelligence agencies.
In its notice, the Department of Homeland Security stated that DeepSeek models present multiple operational threat vectors and could result in the transfer of sensitive U.S. government data to China. The department warned that DeepSeek maintains its servers in China, where the legal system grants state authorities extensive access to private corporate infrastructure.
The federal restriction follows accusations from U.S. officials and domestic AI companies regarding intellectual property practices among Chinese competitors. In an internal memo, White House tech lead Michael Kratsios argued that rival firms in China have engaged in the malicious exploitation of American AI developers.
Major U.S. AI developers, including OpenAI and Anthropic, have also claimed that Chinese competitors such as DeepSeek copied their proprietary models. Those domestic firms are simultaneously navigating regulatory pressure at home, after eight states sued OpenAI and the Federal Trade Commission subpoenaed both companies in mid-October to investigate consumer safety risks.
DeepSeek disrupted the global artificial intelligence sector in January 2025 when it released an open-source model that matched the performance of leading U.S. chatbots at a fraction of their development cost. That efficiency made DeepSeek tools popular among commercial software companies and developers looking for alternatives to closed American models built by OpenAI, Google, and Anthropic.
The competition over artificial intelligence has become a central point of geopolitical friction between Washington and Beijing, running alongside broader economic disputes as European Union leaders also weigh new trade measures against China. Despite the regulatory restrictions, both countries have sought to manage the risks of the technology race. In September 2026, U.S. President Donald Trump and Chinese President Xi Jinping met at the White House and agreed to establish a bilateral communication channel for artificial intelligence incidents.
Why it matters
The prohibition marks a major escalation in federal efforts to insulate government information systems from Chinese artificial intelligence technology. Because DeepSeek's open-source models are widely used by commercial developers due to their low cost and high performance, barring third-party software built on those models from federal contracting could affect a broad range of commercial software vendors that supply the U.S. government.
What's next
Trump and Xi are scheduled to meet for a summit in Shenzhen, China, in November 2026, where artificial intelligence competition is expected to be on the agenda. In Washington, questions remain over how broadly the contracting ban will affect commercial vendors that integrate DeepSeek code, and whether the Department of Defense or U.S. intelligence agencies will issue parallel bans for military and intelligence systems.
Sources · 3 reports
Spotted an error? Write to [email protected]. Tags: Artificial intelligence, China, Department of Homeland Security, Cybersecurity, DeepSeek, White House.