Full Breakdown
DHS Bans Chinese AI Model DeepSeek from Civilian Federal Agencies
By Drooid · · How we work
Core Action: Prohibition of DeepSeek for Non-military Federal Use
The Department of Homeland Security (DHS) issued a notice ordering civilian federal agencies to stop purchasing, using, or retaining any information-technology products or services derived from Hangzhou DeepSeek Artificial Intelligence. The ban applies to all non-military agencies—including Treasury, State, and Justice—and to any technology built with DeepSeek’s models. Defense-related entities such as the Department of Defense and the Office of the Director of National Intelligence are exempt. The order cites the Federal Acquisition Supply Chain Security Act of 2018 as its legal authority and requires agencies to remove existing DeepSeek installations while monitoring for residual risk.
Background & Context
DeepSeek, a Chinese open-weight AI model released publicly in 2025, offers free access and lower operating costs than proprietary alternatives. Its rapid adoption by U.S. startups has drawn scrutiny from officials who warn that the model’s data-storage practices could expose sensitive information to the Chinese government. Earlier legislative attempts, such as Rep. Josh Gottheimer’s (D-New Jersey) bill to bar DeepSeek from federal use, stalled without passage.
Timeline
- June 26 – Secretary of Homeland Security Markwayne Mullin sent a letter ordering the exclusion of DeepSeek products, indicating the ban had been in effect for months.
- Monday – DHS posted the public notice in the Federal Register, making the ban official.
- End of July – The prohibition took effect, requiring agencies to cease all DeepSeek-related activity.
- Last week – DeepSeek released its most recent model version, generating renewed user praise online.
Data & Statistics
- DeepSeek’s debut model cost roughly $6 million (£4.5 million) to train, compared with about $100 million for OpenAI’s GPT-4.
- The free chatbot app ranks among the most downloaded on the Apple App Store; its servers have experienced downtime due to high demand.
- No mass usage of DeepSeek has been reported within the federal government, though a Department of Agriculture employee attempted to access the chatbot in 2023.
Official Statements & Responses
Secretary Mullin warned that DeepSeek’s models present “multiple threat vectors” and that data stored on Chinese servers could be compelled by Chinese national-security laws. “The transfer of the aforementioned information to the [People’s Republic of China]… could produce actionable intelligence on U.S. persons and/or government programs,” he said.
Former President Donald Trump reiterated his administration’s stance on AI competition, stating, “We’re leading now over China and everyone else, and I want to keep it that way.”
White House technology adviser Michael Kratsios, in an internal memo earlier this year, warned that “malicious exploitation” of American AI by rival Chinese firms had been observed, citing concerns that Chinese models may systematically distill U.S. AI research.
The notice also directs agencies to enhance security controls—such as increased traffic monitoring, enhanced logging, and data-exfiltration safeguards—while they phase out DeepSeek components.
What’s Next
Agencies may request a waiver from the ban by submitting a written justification at least 60 days in advance, citing mission-critical needs or other national-security considerations. The DHS expects affected entities to implement enhanced monitoring and logging measures during the transition period. No further legislative action on the issue has been reported.