DeepSeek deep-sixed at all US federal civilian agencies

On Monday, the US Department of Homeland Security (DHS) issued an order prohibiting federal civilian agencies from utilizing technology products from Chinese AI lab DeepSeek, along with removing any existing implementations. The directive encompasses all DeepSeek and DeepSeek-affiliate products, as well as third-party products and services developed using DeepSeek's technology.

Compliance poses challenges as it is unclear which third-party products and services have been developed using DeepSeek's components. It is also uncertain whether "developed using" refers to providing ongoing access to DeepSeek models via API or merely to apps coded with DeepSeek's assistance. The order also raises questions about AI gateways, such as OpenRouter and Vercel AI Gateway, which make DeepSeek's models accessible through APIs.

Federal agencies might inadvertently use DeepSeek models if they have not restricted their AI providers. The DHS spokesperson did not respond to a request for comment. Concerns about DeepSeek's security and privacy have existed since the release of DeepSeek-R1 in early 2025. The order, while long overdue, may benefit American AI companies like Anthropic, as open-weight models pose competition and are popular due to their capabilities and cost-effectiveness.

However, the order's scope is broad, potentially affecting US-hosted versions of DeepSeek's models. Experts argue that while banning DeepSeek's app and API is feasible, the ban on third-party products and services developed using DeepSeek's implementations is impractical, as there is no reliable way to verify adherence to the restriction.