Washington’s stance on the global artificial intelligence race has entered a new, more aggressive phase as US Treasury Secretary Scott Bessent signaled that the administration is prepared to deploy financial sanctions against Chinese AI developers suspected of misappropriating American intellectual property. Speaking on FOX Business’ "Mornings with Maria" on Tuesday, Bessent outlined a policy framework that balances the Trump administration’s stated support for domestic open-source innovation with a "zero-tolerance" approach toward foreign entities that utilize American technological breakthroughs to accelerate their own sovereign AI programs.
The Treasury Secretary’s remarks represent a significant pivot in the ongoing trade and technology dispute between the world’s two largest economies. While previous restrictions have primarily focused on hardware—specifically the high-end semiconductors and lithography equipment required to train large language models (LLMs)—the proposed sanctions would target the software, weights, and training methodologies of the models themselves. This move suggests that the White House views the software layer of the AI stack as a critical frontier of national security and economic sovereignty.
The Framework of Potential Sanctions
Secretary Bessent’s warning was explicit regarding the consequences of intellectual property (IP) theft. "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. He emphasized that while the administration remains a proponent of the open-source movement within the United States—viewing it as a catalyst for American competition—it would not allow that openness to be exploited by strategic adversaries.
The mechanism for such sanctions would likely involve the Department of the Treasury’s Office of Foreign Assets Control (OFAC). By placing Chinese AI firms on the Specially Designated Nationals (SDN) list or the Entity List, the US could effectively freeze those companies out of the dollar-based financial system, prohibit US persons from investing in them, and block their access to American cloud computing resources. This would create a "Silicon Curtain," isolating Chinese developers from the global ecosystem of researchers, capital, and infrastructure that has defined the AI boom of the 2020s.
The Technical Crux: Model Distillation as IP Theft
At the heart of the current dispute is a practice known as "model distillation." This technique involves using the outputs of a highly advanced, "teacher" model (such as OpenAI’s GPT-4 or Anthropic’s Claude 3.5) to train a smaller, more efficient "student" model. By analyzing the logic and responses of the superior system, developers can create high-performing models at a fraction of the original R&D cost.
US-based AI giants have increasingly characterized this practice as a form of sophisticated theft. They argue that their models represent billions of dollars in capital expenditure and years of proprietary research, and that "scraping" their APIs to build competing products violates terms of service and undermines the competitive advantage of American firms.
However, the definition of distillation as "theft" is a subject of intense debate within the global tech community. Critics of the administration’s stance argue that distillation is a standard academic and industrial optimization technique used worldwide. They suggest that labeling it as IP theft could set a dangerous precedent, potentially criminalizing common reverse-engineering practices that have been part of the software industry for decades.
The Rapid Rise of Chinese AI Capabilities
The timing of Bessent’s warning coincides with a period of remarkable progress for Chinese AI labs. For several years, the prevailing narrative in Silicon Valley was that China remained two to three years behind the US due to the lack of access to NVIDIA’s H100 and B200 chips. Recent benchmarks, however, suggest that the gap is narrowing rapidly.
Moonshot AI recently unveiled its Kimi K3 model, which has demonstrated elite performance in coding, mathematical reasoning, and "agentic" tasks—areas where US firms like OpenAI and Anthropic have traditionally held a dominant lead. Similarly, DeepSeek, another prominent Chinese developer, recently released DeepSeek-V3, which rivals top-tier Western models in efficiency and performance while reportedly being trained on a significantly smaller hardware budget.
These advancements have alarmed Washington policymakers. A recent report from Axios suggested that the Trump administration has been weighing even broader restrictions, including potential curbs on the export or public release of open-source model weights if they are deemed to provide a "dual-use" military advantage to China. While some officials have disputed the extent of these proposed restrictions, Bessent’s comments confirm that the administration is actively seeking ways to protect the "moat" around American AI innovation.
Industry Reactions and the "Fair Use" Paradox
The prospect of sanctions has triggered a complex reaction from industry leaders, some of whom have highlighted the apparent contradictions in how AI companies view intellectual property. Microsoft CEO Satya Nadella recently voiced skepticism regarding the aggressive IP claims made by some AI developers. Nadella pointed out a fundamental irony: major AI companies rely heavily on "fair use" doctrines to justify training their models on massive troves of public data—including copyrighted books, news articles, and code—yet they impose highly restrictive terms of service to prevent others from using their own model outputs for training.
This legal tension was further highlighted by recent developments involving Anthropic. A federal judge recently approved a process for Anthropic to begin payments under a $1.5 billion settlement with a group of authors. The lawsuit alleged that the company had illegally downloaded and stored copyrighted books to train its Claude models. Critics argue that if US companies are found to have "stolen" data from authors and creators to build their models, their moral and legal standing to accuse foreign rivals of "stealing" via distillation is significantly weakened.
Clem Delangue, the CEO of Hugging Face—the world’s largest platform for open-source AI—has also weighed in on the debate. Delangue argued that the US might be overestimating the role of distillation in China’s success. He attributed the rapid rise of Chinese AI to the country’s massive pool of engineering talent, strong research culture, and a more collaborative approach to development within their domestic ecosystem. Delangue cautioned that aggressive sanctions could backfire, encouraging China to become entirely self-reliant and potentially leading to a fragmented global AI landscape where Western standards are no longer the default.
A Chronology of Escalation
To understand the weight of Bessent’s comments, it is necessary to view them within the timeline of the broader US-China tech "cold war":
- October 2022: The US Department of Commerce issues sweeping export controls on advanced semiconductors and chip-making equipment, specifically targeting China’s ability to develop high-end AI.
- May 2023: Major US AI labs, including OpenAI and Google, begin implementing stricter API monitoring to detect and block "automated" queries that appear to be used for model distillation.
- October 2023: The Biden administration updates export controls to close loopholes, including restrictions on "gray market" chip sales and access to US cloud services by foreign entities.
- Mid-2024: Chinese models like DeepSeek and Kimi begin topping global benchmarks, leading to calls from some US lawmakers for "software-level" intervention.
- July 2024: Reports emerge of a $1.5 billion settlement involving Anthropic, highlighting the domestic legal vulnerabilities of US AI training practices.
- January 2025 (Present): Treasury Secretary Scott Bessent officially links model distillation and IP theft to the threat of financial sanctions.
Economic and Geopolitical Implications
The imposition of sanctions on Chinese AI developers would have far-reaching implications for global markets. For the Treasury Department, the challenge lies in enforcement. Unlike physical goods, AI models are digital and can be distributed via decentralized networks. Tracking whether a model was "distilled" from an American source requires forensic analysis of model weights and behaviors—a task that is technically difficult and prone to error.
Furthermore, such sanctions could accelerate the "bifurcation" of the global economy. If Chinese firms are blocked from using American AI tools, they will likely accelerate the development of their own hardware (such as Huawei’s Ascend chips) and software frameworks. This would create two distinct tech stacks: one led by the US and its allies, and another led by China, with neutral nations forced to choose between them.
From a financial perspective, sanctions would also deter global venture capital from flowing into any Chinese startup with AI exposure, fearing that a future Treasury designation could render their investments worthless. This "chilling effect" is a deliberate part of the strategy to starve the Chinese AI sector of the capital and expertise it needs to scale.
Analysis: The Future of AI Diplomacy
Secretary Bessent’s comments signal that the US is moving away from a purely defensive posture—blocking inputs like chips—toward an offensive posture that targets the outputs and the economic viability of foreign competitors. This "software-first" approach to national security recognizes that in the age of AI, data and logic are as strategically valuable as oil or steel.
However, the administration faces a delicate balancing act. If the US defines "IP theft" too broadly, it risks stifling the very open-source innovation that Bessent claims to support. If it defines it too narrowly, it may fail to prevent China from "free-riding" on American research.
As the Trump administration settles into its second term, the Treasury’s role in technology policy is likely to expand. The focus on AI is not merely about protecting corporate profits; it is about ensuring that the foundational technology of the 21st century is built upon Western values and economic structures. Whether sanctions will be enough to halt the momentum of Chinese developers remains to be seen, but the "warning shot" fired by Secretary Bessent indicates that the US is ready to use every financial tool at its disposal to maintain its lead in the artificial intelligence revolution.
