US orders government agencies to remove DeepSeek AI tools over security risks

The United States has directed government agencies to cease employing artificial intelligence tools created by Chinese firm DeepSeek, citing security flaws and the danger of sensitive data being transmitted to China. The Department of Homeland Security announced this prohibition on Monday, mandating the removal of DeepSeek products and third-party applications utilizing its technology from government IT systems.

While the department can issue such orders to civilian bodies, it cannot direct the military or intelligence services.

DeepSeek's AI models, according to AFP, contain "various vulnerabilities that can be exploited as threats," according to a document attached to the notice. The department cautioned that using DeepSeek's AI tools could put sensitive information at risk if accessed by US government personnel or contractors, highlighting that DeepSeek's servers are situated in China, a nation with extensive legal authority over private company infrastructure.

The Pentagon refrained from responding to an AFP query for comment, while AFP has reached out to DeepSeek for a statement. DeepSeek, one of China's premier AI laboratories, made headlines in January 2025 by unveiling a high-performance, budget-friendly chatbot that challenged prevailing notions of US supremacy in the sector. The company unveiled its much-anticipated V4 model in April and released an enhanced, quicker, and more lightweight version this month, as Washington and Beijing vie for dominance in the rapidly developing AI industry.