In a significant hardening of the United States’ stance on global technology competition, US Treasury Secretary Scott Bessent has announced that the federal government is prepared to levy sanctions against Chinese artificial intelligence (AI) developers found to be misappropriating intellectual property from American firms. Speaking during an interview on FOX Business’ "Mornings with Maria," Bessent emphasized that while the Trump administration remains a proponent of the open-source movement within the United States, it will not permit foreign entities to gain a competitive edge by illicitly copying proprietary technologies developed by American innovators.
The Treasury Secretary’s remarks underscore a growing concern within the Department of the Treasury and the broader administration regarding the rapid advancement of Chinese AI capabilities. Bessent’s warning specifically targeted the practice of "theft" in the digital domain, asserting that the US government possesses the necessary regulatory and economic tools to penalize overseas companies that utilize stolen data or model architectures to accelerate their own development. "If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft," Bessent stated, framing the issue as one of both economic fairness and national security.
The Escalation of the US-China AI Competition
The warning from the Treasury Department comes at a pivotal moment in the global AI race. For several years, American companies such as OpenAI, Google, and Anthropic have maintained a perceived lead in the development of Large Language Models (LLMs) and generative AI. However, that lead is increasingly being challenged by a new generation of Chinese models that have demonstrated remarkable proficiency in complex tasks.
Among the most notable examples is Moonshot AI’s Kimi K3, which has recently gained significant traction in international benchmarks. The model has shown high performance in coding, agentic reasoning—the ability of an AI to complete multi-step tasks autonomously—and mathematics. Other Chinese entities, including Alibaba’s Qwen series and ByteDance’s Doubao, have similarly closed the performance gap with US-based frontier models. This rapid progress has led US policymakers to question whether such advances are the result of independent innovation or the systematic "distillation" of American research.
The potential for sanctions marks a strategic shift in Washington’s containment policy. Previously, the US focus remained largely on hardware, specifically restricting China’s access to high-end semiconductors and GPU clusters through stringent export controls managed by the Department of Commerce. By targeting the model developers directly, the Treasury Department is signaling an intent to extend these restrictions to the software layer, potentially cutting off Chinese firms from the global financial system or prohibiting US entities from engaging in business with them.
A Chronology of Technological Containment
The current friction is the latest chapter in a multi-year effort by successive US administrations to maintain technological hegemony. The timeline of these efforts illustrates a steady progression from broad trade tariffs to surgical strikes on specific high-tech sectors:
- October 2022: The US Department of Commerce introduces sweeping export controls designed to restrict China’s ability to purchase or manufacture advanced chips used for AI training, specifically targeting NVIDIA’s A100 and H100 series.
- October 2023: The Biden administration expands these controls to include a wider range of chips and the machinery required to produce them, while also introducing an Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence.
- January 2024: US officials begin investigating "know your customer" (KYC) requirements for cloud service providers to prevent foreign actors from using US-based server farms to train models.
- July 2025: Rumors surface regarding the Trump administration’s intent to revisit the classification of AI weights as a protected national asset.
- July 2026: Treasury Secretary Scott Bessent explicitly links intellectual property theft to the implementation of financial sanctions against Chinese AI firms.
This chronology suggests that the US is moving toward a "Small Yard, High Fence" strategy, where the most critical technologies are guarded with increasing intensity.
The Technical Dispute: Model Distillation vs. Innovation
At the heart of the Treasury Secretary’s warning is the controversial practice of "model distillation." In the context of AI development, distillation involves using the outputs of a highly advanced "teacher" model (such as GPT-4o or Claude 3.5 Sonnet) to train a smaller, more efficient "student" model. This technique allows developers to transfer the sophisticated reasoning capabilities of a large model into a system that is cheaper and faster to run.
American AI companies have long argued that this practice, when performed without authorization, constitutes a violation of their terms of service and an infringement on their intellectual property. They contend that because they spent billions of dollars on compute and data acquisition to train the original model, allowing a competitor to "distill" that knowledge for free is a form of industrial espionage.
However, this position is not universally accepted within the industry. Critics of the administration’s proposed sanctions argue that distillation is a standard scientific method used globally to improve model efficiency. They suggest that the US industry is attempting to redefine common development practices as "theft" to maintain a monopoly on high-performance AI.
Microsoft CEO Satya Nadella has been a prominent voice in this debate, albeit from a nuanced perspective. Nadella has previously questioned the consistency of US AI firms that rely heavily on "fair use" doctrines to train their models on public internet data while simultaneously imposing highly restrictive terms that forbid others from using their own model outputs for training. This perceived hypocrisy has created a rift between those who favor strict IP enforcement and those who advocate for a more open ecosystem.
Supporting Data and Market Impact
The stakes for the US economy are considerable. According to data from various market research firms, the global AI market is projected to contribute trillions of dollars to global GDP by 2030. Maintaining a dominant share of this market is a top priority for the US Treasury.
In 2025, Chinese AI startups reportedly raised over $20 billion in venture capital, despite the ongoing trade tensions. Companies like Moonshot AI and Zhipu AI have seen their valuations soar, driven by domestic demand and the increasing sophistication of their models. If the US proceeds with sanctions, it could disrupt the flow of international capital to these firms and limit their ability to collaborate with global researchers.
Conversely, the impact on US companies could also be significant. Many American tech giants rely on global supply chains and international talent. Furthermore, if China retaliates by restricting access to critical raw materials—such as the rare earth elements required for hardware—the cost of developing AI in the US could rise sharply.
Industry Reactions and Legal Precedents
The reaction to Bessent’s comments has been mixed across the Silicon Valley landscape. Clement Delangue, CEO of Hugging Face—the world’s largest repository for open-source AI models—has expressed skepticism regarding the narrative that Chinese progress is solely due to IP theft. Delangue has argued that China’s advances are the result of world-class research teams and a strategic national focus on AI integration. He warned that over-regulation or aggressive sanctions could stifle the very innovation the US seeks to protect.
Adding complexity to the debate is the ongoing legal volatility within the US AI sector itself. Just as the government seeks to protect US firms from foreign theft, those same firms are facing massive internal legal challenges over their own data acquisition practices.
Recently, Anthropic received court approval to begin payments under a landmark $1.5 billion settlement with a coalition of authors and publishers. The settlement followed a judicial finding that Anthropic had illegally downloaded and stored copyrighted books to train its models. This case highlights a legal paradox: the US government is threatening sanctions against foreign entities for "stealing" from companies that have themselves been found liable for large-scale copyright infringement.
Broader Implications and the Future of AI Diplomacy
The threat of sanctions by the US Treasury signals a move toward what some experts call the "Splinternet"—a world where the digital landscape is bifurcated into two distinct ecosystems: one led by the US and its allies, and the other led by China. Such a division would have profound implications for global research collaboration, technical standards, and the safety protocols governing AI.
From a geopolitical perspective, the use of sanctions as a tool of tech policy reflects the integration of economic security and national defense. If the Treasury Department follows through on Bessent’s warnings, it will likely involve:
- Entity List Additions: Placing Chinese AI labs on the Department of Commerce’s Entity List, preventing them from accessing US software or hardware.
- Financial Restrictions: Utilizing the Office of Foreign Assets Control (OFAC) to freeze assets or prohibit US citizens from investing in targeted AI firms.
- Secondary Sanctions: Penalizing third-party countries or companies that continue to facilitate the "theft" of US AI technology.
The Trump administration’s stance, as articulated by Bessent, suggests that the era of "tech-neutral" trade is over. In its place is a regime where software weights and algorithmic architectures are treated with the same level of scrutiny as nuclear secrets or advanced weaponry.
As the US and China continue to navigate this friction, the global community remains watchful. The outcome of this dispute will likely determine the pace of AI development for the next decade and redefine the boundaries of intellectual property in the age of machine learning. For now, the Treasury Department has sent a clear message: the American government views AI not just as a commercial product, but as a strategic asset that it is prepared to defend through the full force of its economic authority.
