Key points
- DHS bans DeepSeek products from federal civilian agencies and future procurements.
- The order cites risks of data transfer to China and potential malware in models.
- Agencies must remove existing DeepSeek implementations and affiliated services.
Civilian federal agencies must purge DeepSeek AI tools from their IT networks. The Department of Homeland Security issued an order barring the use of the Chinese AI model and its affiliates. This directive targets the removal of any existing versions and prohibits future procurement of related products and services.
The ban stems from national security concerns regarding data handling and system integrity. According to CyberScoop, the order cites the Federal Acquisition Supply Chain Act. It mandates that agencies exclude DeepSeek products to prevent potential threats to federal IT infrastructure.
What we know
Secretary of Homeland Security Markwayne Mullin stated that DeepSeek models present multiple threat vectors to civilian federal IT systems. He noted that the models store data on servers based in China. National security laws in China allow the government to request data from domestic companies.
DHS warned that DeepSeek models could contain malware or other security vulnerabilities. The agency also highlighted privacy concerns, noting that the models collect personally identifiable information. Mullin said the prompt-and-response process may result in the transfer of sensitive data to the Chinese government.
Defender's checklist
- Audit all AI models and services currently deployed in your environment.
- Identify any software built using DeepSeek implementations or weights.
- Remove access to DeepSeek-hosted services and affiliated products.
- Review procurement contracts for clauses related to AI supply chain risks.
What analysts watch for
Teams usually examine how AI models interact with local networks and external endpoints. Analysts check whether models send data to unexpected foreign servers during inference. They look for persistence mechanisms in scripts that load model weights. Observers monitor for unusual outbound traffic patterns that might indicate data exfiltration. Security teams verify that local deployments do not silently communicate with remote hosts. They also assess whether third-party libraries used in the model supply chain introduce vulnerabilities.
What to do and how to stay safe: DeepSeek
- Inventory all AI tools and services used by your organization.
- Check if any tools rely on DeepSeek models or affiliated code.
- Remove access to DeepSeek-hosted services until further notice.
- Monitor logs for unexpected data transfers to foreign servers.
Step-by-step guide: How to Implement Antivirus Software Without Breaking Your Stack
General security guidance from the Malware Chronicle newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Which agencies are affected by the DeepSeek ban?
The order applies to all federal civilian agencies, including the Environmental Protection Agency and Federal Communications Commission.
Why is DHS banning DeepSeek?
DHS cites risks of data transfer to China, potential malware, and privacy concerns regarding data collection.
Does the ban cover third-party products?
Yes, the order excludes DeepSeek products and third-party services developed using its implementations.
