DeepSeek AI software banned across US federal civilian agency networks

Federal civilian agencies in the United States must stop buying and purge all software created by Chinese artificial intelligence firm DeepSeek from their networks, under a new directive issued by the Department of Homeland Security.

The mandate extends past DeepSeek's standalone offerings to encompass third-party tools and applications that integrate the startup's underlying technology. While departments are instructed to purge these systems from existing infrastructure and avoid new acquisitions, the directive permits agencies to apply for specific exemptions if necessary.

Officials at the Department of Homeland Security cited national security risks associated with the startup's base in China. The agency cautioned that sensitive data belonging to federal personnel could be exposed to the Chinese government pursuant to domestic legal mandates governing companies operating in that country.

The ban broadens earlier legislative action enacted in December 2025, which restricted the deployment of DeepSeek within intelligence and defense-related IT networks. The latest policy ensures that civilian bodies across the federal government face equivalent prohibitions.

The government in Beijing swiftly criticized the restriction. Chinese Foreign Ministry spokesperson Mao Ning stated during a briefing that Beijing rejects expanding the definition of national security to politicize trade and commercial technology. She added that authorities would act to defend the lawful rights and commercial interests of domestic companies.

DeepSeek gained global prominence in January 2025 following the release of its R1 reasoning model, a low-cost, open-weight system that disrupted the broader sector and pressured rivals such as OpenAI and Meta. The firm has drawn intense attention from American regulators and industry competitors ever since.

American officials and domestic AI creators have voiced mounting concerns regarding how DeepSeek trains its models. A report issued in September accused DeepSeek and several other Chinese firms, including Alibaba and Moonshot AI, of conducting large-scale distillation to pull capabilities from American AI platforms.

Competitors have previously raised similar allegations. OpenAI told lawmakers in Washington that it detected suspected efforts by DeepSeek to distill data from its models. Anthropic also stated it had caught unauthorized harvesting operations aiming to extract outputs from its Claude system to train competing Chinese architectures.

The federal restrictions arrive against a backdrop of widening tech tensions between Washington and Beijing. Meanwhile, reports indicate that China is contemplating its own export controls across 23 artificial intelligence technologies, including specialized chip blueprints and advanced models, to safeguard against strategic risks.