US bans DeepSeek from the IT systems of federal civilian agencies
IA en un minuto newsroom · Editor: Jon Elgezabal
In 30 seconds
The civilian agencies of the US government will have to strip any DeepSeek product from their equipment and will not be able to buy it. The DHS argues that the Chinese tool may send sensitive data to China and that its models and its iOS app have security flaws. The measure leaves national security systems out and allows exceptions requested within 60 days.
The US Department of Homeland Security (DHS) has ordered the federal government's civilian agencies to remove products from DeepSeek, the Chinese artificial intelligence company, from their IT systems and to stop buying them.
The order is dated July 24 and signed by Homeland Security Secretary Markwayne Mullin. The DHS filed it on Monday, October 19, with the Federal Register, the US government's official journal, which publishes it this Tuesday, October 20. It relies on a 2018 law on supply chain security in federal procurement.
The removal covers products from DeepSeek and its affiliated companies, as well as third-party products and services developed with DeepSeek implementations. Both measures apply from the moment the order was issued, although each agency may request an exception in writing within 60 days, to be reviewed by the cybersecurity agency CISA. The agencies' national security systems and their compartmented information systems are left out, as they fall outside the DHS's remit.
According to the DHS, DeepSeek's apps work through prompts and answers that require handing information to the model, so if government employees or vendors use them, sensitive data may end up in China. The order notes that the company's servers are in China and that its legal system gives the government broad powers to access the infrastructure of private companies. That information, alone or combined with public data, could be used to obtain intelligence on US persons and government programs. The DHS also warns that the models themselves may contain malware or other flaws, and cites independent research by the firm NowSecure that found several critical vulnerabilities in the DeepSeek iOS app.
Until now, restrictions were partial. In January 2025, the Pentagon cut off access to the DeepSeek chatbot through its Defense Information Systems Agency, and several US states have limited its use, according to Bloomberg Línea. In April 2025, a House committee on China described the company's ties to the Beijing government as significant. The order comes out a few days after DeepSeek released its latest model, V4.1-Flash-1017, as Engadget notes.
Why it matters · analysis and opinion
What sets this order apart from earlier bans is its reach: it does not stop at the DeepSeek app, but extends to any third-party product built on its models. That forces agencies, and the vendors that sell them software, to know which model sits inside each service they use. Companies that work for the US federal government will have to be able to prove it if they want to keep doing so. Outside the United States it has no legal effect, but it reinforces the idea that the origin of an AI model is now treated as a security risk in public procurement. It is also one more step in the race between Washington and Beijing to lead artificial intelligence.
Official source: US Department of Homeland Security (Federal Register) · Written with the help of AI: how we make the news
