The United States has formally shut the door on DeepSeek inside the federal civilian government. The Department of Homeland Security has issued a binding supply-chain directive requiring federal civilian agencies to cease using DeepSeek AI products and to remove third-party tools developed with its implementations.

The order reaches beyond the DeepSeek app itself: software built on top of the Chinese AI company's models or code is also in scope, which means agencies — and the contractors and vendors who supply them — must audit where DeepSeek-derived components sit inside their software stacks.

The directive is the latest step in Washington's widening effort to keep Chinese-developed AI out of sensitive government systems, on supply-chain and data-security grounds. For enterprise vendors, the practical consequence is a new compliance burden: proving the provenance of models and components before software can be sold into civilian agencies.

DeepSeek rose to global prominence with capable open models released at strikingly low reported cost, and its tools spread quickly through developer ecosystems — which is exactly what makes a “derived implementations” ban harder to execute than a simple app block.

Agencies now face the unglamorous work of inventorying their AI usage and pulling flagged components, while vendors that can document a clean supply chain may find the directive becomes a selling point.