Federal civilian agencies have four days of paperwork ahead of them. A Department of Homeland Security directive issued Tuesday tells them to stop purchasing products built by Chinese AI developer DeepSeek and to strip existing installations from government IT systems.
The scope runs further than DeepSeek’s own chatbot and API. Third-party tools built on top of the company’s models fall under the same order, though agencies can apply for exceptions.
DHS grounded the decision in data handling. Chinese law could give Beijing’s authorities a route to information that US government users hand to DeepSeek systems, the directive argues.
A ban that reaches into the supply chain
That supply-chain language is the part agencies will find hardest to execute. Open-weight DeepSeek models have been downloaded and fine-tuned by a wide field of startups and internal engineering teams, which means the directive’s target set is bigger than the vendor list suggests.
Beijing pushed back quickly. Foreign Ministry spokesperson Mao Ning told reporters the move was political and would ricochet through global supply chains.
The order lands in a widening technology split. DeepSeek built its reputation on cheap open-weight models that undercut US labs on price, and American officials have spent much of the year closing doors around that distribution channel.
What agencies actually lose is less clear. Several had already limited DeepSeek use following earlier restrictions, and the practical effect may be procurement friction rather than a sudden outage. Exceptions keep a door ajar for work the government decides it cannot replace.
For enterprises watching from the private sector, the message is narrower than it looks. The directive governs federal civilian systems, not commercial deployments. But it sets a template that state governments and regulated industries have followed before.
