US bans DeepSeek AI tools across government agencies over security fears

The United States government has ordered its agencies to halt the use of artificial intelligence tools from Chinese startup DeepSeek, amid escalating tensions in the rapidly evolving AI sector. A notice issued by the Department of Homeland Security mandates the removal of DeepSeek-developed products and tools created using DeepSeek from IT devices, citing security concerns. The department can issue directives to civilian agencies, but not to the military or intelligence services.

DeepSeek's AI models are said to pose various threats, with the potential for sensitive data to be transferred to China if utilized by US government employees or vendors. The warning further notes that DeepSeek's servers are based in China, a country with extensive jurisdiction to access private company infrastructure under its legal system. Attempts to reach DeepSeek for comment have been unsuccessful.

Founded as one of China's leading AI laboratories, DeepSeek made waves in January 2025 by introducing a cost-effective, high-performance chatbot, challenging prevailing perceptions of US supremacy in the field. The company subsequently launched a highly anticipated new model, V4, in April, followed by an updated, more efficient version in the same month. The US has accused DeepSeek and other prominent Chinese AI firms of engaging in the systematic theft of advanced model capabilities from American competitors.

Despite their intense rivalry, China and the United States agreed in September to establish a communication channel for addressing AI-related incidents. This development is anticipated to be revisited when the nations' leaders convene at a summit in Shenzhen, China, next month.