US Homeland Security orders civilian agencies to remove DeepSeek and bans purchases

The US Department of Homeland Security ordered federal civilian agencies to remove DeepSeek products from their information systems and banned their purchase, according to a notice reported by Russian outlets on 19 October. RIA Novosti quoted the notice as saying the Secretary of Homeland Security issued an order covering all products of DeepSeek and its affiliated organisations, as well as products and services of third-party developers created using DeepSeek technology. Civilian authorities may no longer buy products and services of Hangzhou DeepSeek Artificial Intelligence Co., Ltd. [ 1 , 4 , 5 ]

Kommersant reported that the notice was in the Public Inspection section for preliminary review, with official publication in the Federal Register planned for 20 October. It said the department described DeepSeek models as presenting "multiple threat vectors" and that their use could lead to the transfer of sensitive data to China if used by US government employees or vendors. The models themselves may contain malware or other security vulnerabilities, the department said. 3DNews added that, according to the department, use of the models could allow detailed intelligence to be collected on government agencies or US citizens. [ 4 , 6 ]

iXBT reported that the department explained the decision by the risk that Chinese authorities could access sensitive data of US officials, since Chinese law may allow this. According to iXBT, the ban covers new purchases and already installed systems, and government bodies will be able to request exceptions. It said the measures extend restrictions introduced by a December 2025 law affecting DeepSeek in US intelligence national security systems. 3DNews reported that the department had told federal agencies in July to monitor information flows linked to government websites because of DeepSeek, and that the Commerce Department and the Navy had earlier banned DeepSeek on work devices, with New York, Tennessee and Virginia following. [ 6 , 10 ]

CNews, citing The Register, said agencies may face problems complying, because it is not always possible to determine whether DeepSeek tools were used in developing an application. It said the phrase "developed using" is unclear, and that the order's effect on API aggregators such as OpenRouter and Vercel AI Gateway remains open. [ 8 ]

Mao Ning, a spokeswoman for China's Foreign Ministry, said on 20 October that Beijing consistently opposes abstracting the concept of national security and politicising scientific, technological and trade issues, TASS reported. iXBT added that she said China intends to protect the legitimate interests of Chinese companies, and cited Bloomberg as saying China is considering export restrictions on 23 AI-related technologies. Vedomosti recalled that in February OpenAI warned US lawmakers that DeepSeek uses unfair methods to extract results from leading American models to train the next generation of its R1 chatbot. [ 5 , 7 , 10 ]