The U.S. Department of Homeland Security (DHS) issued a directive Monday, ordering every federal agency to cease using AI models from Chinese startup DeepSeek and to uninstall them from all IT equipment. The notice warned that DeepSeek’s servers, housed in China, could expose sensitive data to the Chinese government, citing the country’s broad legal authority over foreign company infrastructure. DHS can direct civilian agencies, but the ban does not extend to the Department of Defense or intelligence services—DOD has yet to comment.
DeepSeek, a leading Chinese AI lab, stunned the industry last January with a low‑cost, high‑performance chatbot that challenged U.S. dominance. Its V4 model, released in April, and a lightweight update this month have pushed the envelope on speed and cost, drawing sharp criticism from Washington.
Washington has long accused DeepSeek and other Chinese firms of stealing U.S. AI technology. In September, the two countries opened a communication channel for AI incidents, a partnership that will likely be revisited at the upcoming summit in Shenzhen.
For the contractors, the ban means a sudden halt to ongoing projects. One federal data‑science team at the National Institute of Standards and Technology, which had been testing DeepSeek's V4 for anomaly detection, must now pivot to a domestic model, adding weeks to their delivery schedule.