Editorial illustration for US Orders Federal Agencies to Remove DeepSeek AI Over Model Distillation
US Bans DeepSeek AI From Federal Agencies Over Security
US Orders Federal Agencies to Remove DeepSeek AI Over Model Distillation
The Department of Homeland Security has told federal civilian agencies to stop buying anything built with DeepSeek's technology and to strip existing versions off government systems. The order, issued this week, doesn't stop at DeepSeek's own chatbot. It reaches into third-party software and services that use the Chinese startup's models underneath, though agencies can apply for exceptions.
DHS says the problem is data. Under Chinese law, officials warn, information from US government users running through DeepSeek's systems could end up accessible to authorities in Beijing. That risk, the department argues, justifies pulling the technology from federal networks entirely rather than restricting it case by case.
The timing matters. DeepSeek rattled the AI industry last year by undercutting OpenAI and Meta on price while releasing open-weight models that performed close to their much more expensive American counterparts. Washington had already moved against the company once, when a December 2025 law barred its use in national security systems tied to US intelligence agencies. This new DHS order widens that restriction to cover the rest of the federal government, new purchases and systems already in place alike.
The U.S. has ordered federal civilian agencies to stop purchasing services and products developed by Chinese AI startup DeepSeek and remove them from government IT systems over national security concerns tied to the Chinese AI startup’s operations in China, according to the Department of Homeland Security.
Why this matters
A federal procurement ban is a blunt instrument, but this one has teeth: it doesn't just bar DeepSeek's app, it reaches into any third-party tool built on top of its models. For developers and founders, that's the real signal. If you've quietly wired DeepSeek or similar distilled models into a product pipeline, you now have a compliance problem the moment a government client or partner asks about provenance.
The distillation allegations matter too. OpenAI and Anthropic have both flagged concerns that DeepSeek trained on outputs from their systems, and DHS's September report names five more Chinese firms, including Alibaba and Moonshot AI, doing the same. That's a direct challenge to how "open" weights and cheap fine-tuning actually get built.
Expect more scrutiny of supply chains for AI components, not just the flashy foundation models. For researchers, it's worth watching whether this becomes a broader US pattern, like the chip restrictions, or stays narrowly about federal purchasing. Either way, the line between "using a model" and "owning its lineage" just got a lot more consequential.
Related Reading
Business & Startups Nordic pilot adds Gemini for Education, NotebookLM to boost AI literacy
Business & Startups Kling launches Video O1, all-in-one model with MVL bridge using transformer
Business & Startups DeepSeek Seeks More Capital Weeks After USD 7B Funding Round
From Archives Enterprise AI Prototypes Break in Production as Hallucinations Persist
From Archives Anthropic co-founder says singularity underway amid AI security focus
Common Questions Answered
Why did the Department of Homeland Security order federal agencies to remove DeepSeek AI from government systems?
The DHS cited national security concerns related to data protection under Chinese law, warning that information from US government users could be subject to access by Chinese authorities. The order specifically targets DeepSeek's model distillation practices and the potential risks associated with storing government data on systems using Chinese AI technology.
Does the federal procurement ban only apply to DeepSeek's own chatbot or does it extend to other products?
The ban extends beyond DeepSeek's own chatbot to include any third-party software and services that use DeepSeek's models underneath. However, federal agencies can apply for exceptions to the order if they have specific operational needs.
What compliance challenges does the DeepSeek ban create for developers and software companies?
Developers and founders who have integrated DeepSeek or similar distilled models into their product pipelines now face compliance problems when government clients or partners inquire about the provenance of their AI technology. This procurement ban creates a significant barrier for any software company seeking to do business with federal agencies that relies on DeepSeek's underlying models.
What is model distillation and why is it a concern in the DeepSeek ban?
Model distillation refers to the process of extracting knowledge from a larger AI model to create a smaller, more efficient version. The distillation allegations against DeepSeek are part of the national security concerns, as they relate to how the company's AI technology was developed and the potential implications for data security.
Further Reading
- U.S. bans DeepSeek from government agency systems amid AI rift with China - CNBC
- DeepSeek barred from government agency IT systems - POLITICO
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning - arXiv
- Improved Methods for Model Pruning and Distillation - arXiv
- Quantifying the Capability Boundary of DeepSeek Models: An Application-Driven Performance Analysis - arXiv
