The U.S. Department of Homeland Security has ordered civilian federal agencies to remove DeepSeek products and third-party services built on the Chinese AI model from government IT systems. The directive also bars agencies from acquiring new DeepSeek-based products. DHS cited concerns that sensitive information could be transferred to China and that the models or associated services could contain vulnerabilities that create security risks for federal networks.
The ban applies across civilian agencies
The directive is broader than earlier agency-by-agency restrictions. It covers civilian government IT systems and extends not only to DeepSeek-branded tools but also to third-party services that rely on the model.
DHS cited multiple threat vectors
The department said use of DeepSeek could expose sensitive government or personal information, generate intelligence useful to foreign actors or introduce malware and security vulnerabilities into federal environments.
Data transfer is the core concern
Generative AI systems can receive prompts containing confidential material, internal documents or code. Security officials worry that data entered into foreign-controlled services could be retained or accessed outside the United States.
Several agencies had already acted
The Commerce Department and U.S. Navy had previously restricted DeepSeek on government devices. Multiple states including New York, Tennessee and Virginia have also imposed limits.
The policy fits a wider China-tech crackdown
The Trump administration has tightened restrictions across telecoms, drones, routers and other technology tied to Chinese companies. DeepSeek is now being treated as part of the same national-security framework.
The FCC is preparing further restrictions
U.S. regulators are considering new rules that would bar Chinese labs from testing certain electronics destined for the American market. The FCC has also moved against imports of some Chinese drones and consumer routers.
Third-party services face a compliance challenge
Software companies that incorporated DeepSeek into products sold to government agencies may now have to replace the model or redesign their services. That could create short-term costs for vendors and procurement teams.
Government AI policy is becoming model-specific
Federal agencies are increasingly distinguishing between AI systems based on where models are developed, how data is handled and who controls the underlying infrastructure. The DeepSeek action shows that AI procurement is now a security-policy issue as much as a software choice.
Open-source availability complicates enforcement
Some DeepSeek models can be run locally rather than accessed through a hosted service. That raises technical questions about whether agencies should block only cloud endpoints or also prohibit local deployment of model weights.
Monitoring web traffic will become part of compliance
A DHS memo made public Monday instructed agencies to watch traffic flows for signs of DeepSeek use or compromise. That suggests enforcement will depend partly on network monitoring rather than simple software inventories.
The decision could influence private companies
Federal security rules often become benchmarks for contractors and regulated industries. Companies handling government data may adopt similar restrictions even if they are not directly required to do so.
Sources
- Reuters, Oct. 19, 2026
- U.S. Department of Homeland Security notice
