US government bans federal agencies from using DeepSeek AI tools

Federal civilian employees and vendors must now scrub all artificial intelligence products created by Chinese startup DeepSeek from their IT devices and systems. The Trump administration directive, made public on Monday in the Federal Register by the Department of Homeland Security, forces civilian agencies to remove any third-party technology developed using DeepSeek models and bans all future purchases of the firm's services.

What threat vectors do US officials see in DeepSeek?

The Department of Homeland Security issued a document attached to Monday's notice warning that DeepSeek's models present multiple threat vectors and that using them may result in the transfer of sensitive data to [China] if used by US government employees or vendors. The directive points out that the company's servers sit in China, where the domestic legal system grants Beijing extensive authority to access private company infrastructure. Furthermore, the agency cited open-source research identifying multiple critical vulnerabilities in DeepSeek's iOS app, raising fears of malware and security flaws.

That leaves executive agencies scrambling to comply with a sweeping directive that covers departments such as Treasury, State, and Justice. While the order allows executive agencies to request exceptions or compliance delays by submitting a written justification to the Department of Homeland Security within 60 days of issuance, the military and intelligence services operate under separate authorities. The Department of Defense did not respond to an AFP request for comment, though the Defense Information Systems Agency previously moved to block access to DeepSeek models back in January 2025.

How has Beijing responded to the directive?

China hit back sharply at the restrictions, accusing Washington of weaponizing national security to handicap foreign competitors. Chinese Foreign Ministry spokesperson Mao Ning told reporters on Tuesday that China consistently opposes the overstretching of the concept of national security and the politicizing of science, technology, economic and trade issues. Mao added that Beijing would resolutely safeguard the legitimate rights and interests of Chinese companies.

The tech trade friction forms part of a broader race for dominance in artificial intelligence between the two superpowers. The United States has accused DeepSeek and other top Chinese AI firms of systematically stealing the capabilities of cutting-edge models developed by American rivals. DeepSeek, recognized as one of China's top AI labs, caught the industry off guard in January 2025 by launching a low-cost, high-performance chatbot that challenged long-held assumptions of American technological supremacy. The startup subsequently released a long-awaited new model, V4, in April, and earlier in October updated a faster, lightweight version of V4.

How did federal restrictions build up over time?

1. January 2025: DeepSeek launches a low-cost, high-performance chatbot, while the Defense Information Systems Agency moves to block access to its models.
2. March 2025: U.S. Commerce Department bureaus inform staff that DeepSeek is banned on government devices.
3. July 24, 2025: The Department of Homeland Security originally issues the civilian agency order.
4. September 2025: China and the United States agree to establish a communication channel for AI incidents when their leaders meet at the White House.
5. December 2025: Congress restricts the use of DeepSeek in intelligence community systems.
6. April 2026: DeepSeek releases its V4 model.
7. Earlier in October 2026: The firm updates a faster, lightweight version of V4.
8. Monday, October 20, 2026: The Department of Homeland Security order is officially made public in the Federal Register.

What should affected federal IT users check and do?

Federal civilian employees, contractors, and agency IT administrators must immediately audit their hardware and software inventories to identify and excise prohibited applications. Specifically, personnel must take the following steps:

• Check all government-issued IT devices, mobile phones, and workstations for any direct applications or software products developed by DeepSeek.
• Examine third-party software tools and vendor platforms to determine whether their underlying architecture or generative AI components rely on DeepSeek technology.
• Ensure that no future procurement requests or purchase orders include products or services tied to the startup.
• Submit any necessary written justifications for compliance exceptions or delays through official channels to the Department of Homeland Security within the designated 60-day window.

Will this diplomatic rift derail upcoming leadership summits?

Despite the escalating technological decoupling, both governments have left diplomatic doors open for high-level talks. Following an agreement in September to set up a dedicated communication channel for artificial intelligence incidents when their leaders Xi Jinping and Donald Trump met at the White House, the broader digital rivalry is scheduled to top the agenda when President Donald Trump and Chinese President Xi Jinping meet next month at a summit in the Chinese city of Shenzhen. DeepSeek itself did not immediately respond to requests for comment from media outlets, leaving its leadership's direct response unvoiced as the diplomatic showdown approaches.

The widening federal ban exposes the fragile reality of tech governance, where a foreign startup's rapid code iteration can trigger multi-agency purges months before diplomats sit down to talk. That leaves civilian administrators enforcing rigid software blacklists while the White House prepares to negotiate directly with Beijing over the future of global AI borders in November.