The US government has ordered civilian agencies to remove DeepSeek’s AI technology from their systems, citing concerns that sensitive information could be transferred to China.
The US government has ordered civilian agencies to remove artificial intelligence (AI) technology developed by Chinese startup DeepSeek from their systems, escalating concerns over the security of Chinese AI tools.
The Department of Homeland Security (DHS) issued the order on Monday (19.10.26), banning DeepSeek products and technology created using its models from civilian government IT systems.
The directive also applies to third-party tools built with DeepSeek technology.
It took effect at the end of July but was made public this week.
The DHS said DeepSeek’s models presented “multiple threat vectors”, warning that their use by government employees or contractors “may result in the transfer of sensitive data to [China] if used by US government employees or vendors”.
It also highlighted the location of DeepSeek’s servers, saying China’s legal system gives the government “extensive authority” to access private company infrastructure.
The ban applies to civilian agencies including the departments of Treasury, State and Justice.
It does not cover agencies within the Department of Defense, while the DHS does not have authority over military or intelligence services.
The move comes as the US and China compete for leadership in AI, with President Donald Trump repeatedly identifying China as America’s key rival in the development of the technology.
DeepSeek caused a major stir in January 2025 when it released a low-cost, high-performance chatbot that challenged assumptions about the enormous computing resources needed to develop advanced AI.
Its models are open source, meaning they can be accessed and adapted more cheaply than many proprietary systems.
US technology companies including Snowflake, Vercel and OpenRouter have incorporated DeepSeek models into their products.
The startup’s low development costs have also attracted significant attention.
DeepSeek previously said it spent about $6 million (£4.5 million) training its first model, compared with estimates of roughly $100 million for OpenAI’s GPT-4.
US AI companies have accused Chinese rivals of copying or extracting capabilities from their models.
OpenAI and Anthropic have both previously raised concerns about Chinese firms using their technology.
Despite the tensions, the US and China agreed in September to establish a “communication channel” for AI-related incidents.
Donald Trump and Chinese President Xi Jinping are expected to discuss AI again at a summit in Shenzhen in November.
The latest DeepSeek ban highlights the growing tension between the technology’s widespread availability and governments’ concerns about where AI systems are developed, hosted and controlled.
US orders DeepSeek AI removed from government systems
