The US Department of Homeland Security (DHS) has ordered the removal of artificial intelligence technology from the Chinese firm DeepSeek across most civilian government agencies. According to MyJoyOnline, the ban covers both DeepSeek's own models and third-party tools built using them, taking effect at the end of July but made public on Monday. The order specifically targets civilian departments such as the Treasury, State, and Justice departments, while excluding agencies within the Department of Defence. Punch reported that the DHS stated DeepSeek's models present multiple threat vectors, warning that their use could transfer sensitive data to China if accessed by US government employees or vendors.
Beijing responded on Tuesday by opposing the ban, with the foreign ministry describing the move as an overstretch of the concept of national security. Punch reported that ministry spokesperson Mao Ning told a news conference that China has consistently opposed the politicisation of science, technology, and trade issues. The US government's rationale, as detailed by Khaleej Times, includes the fact that DeepSeek's servers are located in China, where the legal system grants the government extensive authority to access private company infrastructure. Asharq Al-Awsat noted that the Department of Defence did not respond to requests for comment regarding the exclusion of military and intelligence services from the ban.
The ban comes amid a broader race for dominance in the AI sector, with the US accusing Chinese firms of systematically stealing capabilities from American rivals. MyJoyOnline noted that DeepSeek has been seen as a disruptive force because its open-source models are free to access and significantly less expensive to use than closed models from US firms. The company's debut model in 2025 was reported to cost approximately $6 million to train, a fraction of the cost for comparable US models. Punch reported that DeepSeek released a new model, V4, in April and updated a faster, more lightweight version this month. The topic is expected to arise when the US and Chinese leaders meet at a summit in Shenzhen next month, following their agreement in September to set up a communication channel for AI incidents.