Trump administration considers rule for disclosing Chinese AI model use
The administration is weighing a disclosure requirement for US companies using Chinese AI models as part of a broader national security strategy.
The short version
- The proposed rule would require public disclosure of Chinese AI model use without banning the technology.
- Officials are also considering procurement restrictions and adding Chinese AI labs to the Entity List.
- Concerns center on 'distillation,' where Chinese firms learn from the outputs of stronger US models.
- The Commerce Department recently tightened export controls on AI chips to limit Chinese access to compute resources.
Why it matters
For businesses, this could mean new transparency requirements and potential licensing hurdles for using international AI tools. It signals a shift toward more active government oversight of the global AI supply chain.
The disclosure proposal
The Trump administration is currently evaluating a new rule that would force American companies to publicly state whether they utilize Chinese AI models. This specific proposal is being viewed as a relatively mild regulatory tool because it does not actually prohibit the use of such technology. Instead, it focuses on transparency, requiring firms to be open about their reliance on Chinese-made systems.
This approach aligns with a broader internal discussion regarding how the government should respond to Chinese AI development. Reports suggest that officials are prioritizing voluntary cooperation and targeted restrictions over the implementation of sweeping, nationwide bans. By focusing on disclosure, the administration can monitor the flow of technology without immediately disrupting the operations of domestic businesses.
The debate over these rules comes at a time when Washington is looking for ways to manage the influence of Chinese AI labs. While disclosure is the current focus, the administration is also exploring more aggressive measures. These include procurement restrictions and the possibility of placing specific Chinese AI entities on the Entity List to limit their access to US technology.
How distillation works
A major driver behind these policy discussions is a process known as distillation. This is a training technique where one AI model learns to perform tasks by studying the outputs generated by another, typically much stronger, model. In this scenario, Chinese firms may be using the outputs of US models to improve their own systems without needing to build the underlying technology from scratch.
Senior officials have expressed significant concern that this practice could pose a national security risk. Treasury Secretary Scott Bessent and White House science adviser Michael Kratsios have both issued warnings that the distillation of US models could lead to potential sanctions. They worry that the intellectual property and capabilities of American AI are being extracted and repurposed by foreign competitors.
The government has already begun taking action to address these specific concerns in certain sectors. Under provisions in the 2026 NDAA, the Department of Defense, the Director of National Intelligence, and the Department of Homeland Security have mandated that specific Chinese models, such as DeepSeek, be excluded from government systems. These actions show a clear intent to wall off sensitive government infrastructure from Chinese AI influence.
Hardware and export controls
While the software side of the AI debate focuses on models and distillation, the hardware side focuses on the physical chips required to run them. On November 16, 2026, the Commerce Department issued a rule designed to tighten export controls on AI chips. This move is intended to limit the ability of Chinese entities to exploit American compute resources for their own development.
The new rule specifically targets overseas data centers that provide significant compute resources to firms based in China. These facilities are now required to obtain a license within 90 days to continue providing those resources. This creates a significant hurdle for the infrastructure that powers large-scale AI training and deployment, potentially slowing the pace of Chinese technological advancement.
These hardware restrictions complement the potential software disclosure rules by creating a two-pronged strategy. One side aims to limit the physical means of production through export controls, while the other seeks to monitor the usage of existing models. Together, these policies represent a coordinated effort to manage the geopolitical competition surrounding artificial intelligence and the markets that support it.
Internal policy divides
The path forward for these regulations is not without internal friction within the administration. While the disclosure rule is being considered, other more severe options have been discussed but not yet fully executed. These include adding more companies to the Entity List and creating executive orders that would tie liability for security breaches to the use of certain technologies.
There is a visible divide between different factions of officials regarding how to proceed. National security hawks argue for firm, uncompromising lines against Chinese AI to protect American interests. They believe that any lack of restriction could allow foreign adversaries to gain a significant technological advantage over the United States.
On the other side of the debate, some officials worry that heavy-handed restrictions could have unintended consequences for the domestic economy. They argue that overly strict rules might slow down American innovation and hurt the competitiveness of US companies in the global market. This tension between security and economic growth continues to shape the administration's deliberation on the final rules for AI disclosure and usage.
What to watch
- Whether the administration moves from disclosure to a formal ban on specific open-weight models.
- How many Chinese AI labs are eventually added to the Entity List.
- The impact of the 90-day licensing window on overseas data center operations.
Try it
Check if your company's AI policy includes a registry of the specific models and providers used in your daily workflows.
Sources and links
First reported by Crypto Briefing. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.