U.S. Bans DeepSeek, China's Commerce Ministry Demands Reversal of Restrictions
nashnova research
The U.S. Department of Homeland Security ordered federal agencies to purge all DeepSeek products and ban future procurement. China's Commerce Ministry fired back, demanding the U.S. rescind the restrictions — the AI control battle is spreading from chips to models.
What exactly does this ban cover?
DHS ordered federal civilian agencies to remove every DeepSeek product from their IT systems and block all future purchases.
The stated concern: China's government could use DeepSeek to access sensitive U.S. data and "generate actionable intelligence."
This means → DeepSeek isn't just restricted — it is being purged entirely from U.S. federal infrastructure, echoing the playbook used against Huawei equipment.
Why can't open-source dodge the ban?
The order reaches beyond DeepSeek's own products to cover third-party companies that built on its open-source technology — broader than most observers expected.
DeepSeek, Moonshot AI, Alibaba, and other Chinese AI firms let developers download open-source models for free and build on top of them — a model that made them attractive in the U.S. commercial market.
In plain terms = the old assumption was "open-source means safe." This ban says: if the underlying technology is DeepSeek's, no matter who packaged it, the federal system can't touch it.
How did China respond?
China's Commerce Ministry called the U.S. allegations "groundless" and a pretext for monopolizing the AI industry, demanding Washington "rescind" the restrictions and warning that "China will take resolute countermeasures."
Foreign Ministry spokesperson Mao Ning said Tuesday that China opposes the over-extension of national-security concepts and the politicization of technology and trade.
This reflects a clear escalation in Beijing's language — from "expressing concern" to "resolute countermeasures" — signaling that reciprocal action may follow.
What is the "copying" accusation about?
The FBI and other U.S. agencies last month accused DeepSeek and other Chinese AI firms of "large-scale, targeted" copying of frontier U.S. AI models.
The FBI added that the effort was "very likely" conducted with the knowledge of China's government. Beijing denied the allegations.
This means → Washington is building a dual narrative — national security plus intellectual property — framing Chinese AI progress not just as a data risk but as stolen technology.
Where does this standoff go from here?
The U.S. has spent years restricting advanced chip exports to China to curb its AI progress; DeepSeek upended that logic last year by building a top-tier AI system for under $6 million in training costs.
This spring, DeepSeek disclosed that its new model runs on Huawei chips — seen as a milestone in reducing China's dependence on foreign technology.
Last week, China's Commerce Ministry and science ministry proposed export controls and licensing requirements on certain AI technologies, mirroring Washington's control logic — though already-released open-source models are exempt for now.
In plain terms = the control battle is escalating from "who can buy which chips" to "who can use which models" — the scope and intensity of the next countermeasures are the key variable for where this standoff heads.
市场有风险，内容仅供研究参考，不构成投资建议。