Federal civilian agencies must now purge DeepSeek AI from their IT infrastructure per a Department of Homeland Security order. This mandate also prohibits the future purchase of any products or services associated with the Chinese AI firm.
The Federal Acquisition Supply Chain Security Act of 2018 and the DeepSeek Ban
The Department of Homeland Security issued this directive under the authority of the Federal Acquisition Supply Chain Security Act of 2018. According to the report, the order requires agencies to remove all products affiliated with DeepSeek, including any third-party services that were developed using the Chinese model. The move follows a formal review and recommendation from the Federal Acquisition Security Council.
The Department of Homeland Security cited multiple "threat vectors" as the primary motivation for the ban. Specifically, the agency expressed concerns that government employees or vendors using DeepSeek could inadvertently transfer sensitive U.S. data to the People's Republic of China.
Malware Risks and the Monitoring of Government Website Traffic
Beyond data privacy, the Department of Homeland Security warned that DeepSeek models could be used to generate actionable intelligence on U.S. government agencies or individual U.S. persons. as the report says,the department believes the models themselves may contain security vulnerabilities or embedded malware.
To mitigate these risks, a July memo made public on Monday instructed federal agencies to closely monitor traffic flowing in and out of government websites. This surveillance is intended to identify and address potential security breaches linked to the use of DeepSeek tools.
How DeepSeek's Open-Weight Models Shook Nvidia and Microsoft
DeepSeek entered the global market early last year, offering a cost-effective alternative to the closed-model systems built by U.S. giants like OpenAI, Google, and Anthropic. the sudden emergence of these cheaper models caused significant volatility in U.S. tech stocks, specifically impacting companies like Nvidia, Microsoft, and Tesla.
The primary appeal of DeepSeek lies in its "open-weight" architecture. Unlike closed models, open-weight models make their trained parameters, source code, and training data public, allowing users to customize the AI more easily. While the report notes that DeepSeek's performance has historically lagged behind top U.S. models, that gap has narrowed over the last year.
From the U.S. Navy to New York State's Privacy Concerns
The broad federal ban follows a series of smaller, targeted restrictions. The U.S. Navy had previously directed its members to avoid the model, and the Department of Commerce banned DeepSeek from government devices last year due to security concerns.
State-level governments have mirrored this caution. New York State banned the AI from its networks, citing censorship risks and data privacy concerns. Similar restrictive measures have been implemented in Tennessee and Virginia, reflecting a wider trend under the Trump administration to curb the influence of Chinese technology firms.
The October 29 FCC Vote and the Missing DHS Deadlines
The crackdown on Chinese tech extends beyond AI. the Federal Communications Commission (FCC) is scheduled to vote on October 29 to bar Chinese laboratories from testing electronic devices—such as computers and smartphones—intended for the U.S. market.. This is a significant shift, given that 82 percent of U.S. electronics were tested in China last year.
Despite the urgency of the DHS order, several critical details remain missing. The Department of Homeland Security has not provided a specific deadline for agencies to complete the removal of DeepSeek, nor has it elaborated on the specific technical security concerns that triggered the decision. Furthermore, the report does not mention if the Chinese government or DeepSeek has responded to these allegations.
Comments 0