United States Treasury Secretary Scott Bessent has issued a stark warning to the Chinese technology sector, indicating that the Trump administration is prepared to levy sanctions against artificial intelligence companies found to be misappropriating intellectual property from American developers. Speaking during an interview on FOX Business’ "Mornings with Maria" on Tuesday, Bessent articulated a policy stance that seeks to balance the promotion of open-source innovation with the rigorous protection of domestic technological breakthroughs. The Secretary emphasized that while the administration remains a proponent of the open-source movement, it will not remain passive if foreign entities leverage American innovations to gain a competitive edge through illicit means.
"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, signaling a potential shift in the federal government’s approach to managing the escalating technological rivalry between Washington and Beijing. This declaration marks a significant moment in the ongoing trade and technology dispute, moving beyond the restriction of hardware components like semiconductors to the direct targeting of software, algorithms, and the companies that develop them.
The Context of Rising Chinese AI Sophistication
The timing of Bessent’s warning is not coincidental. It arrives at a juncture where Chinese AI models are demonstrating unprecedented capabilities, narrowing the perceived gap between Silicon Valley and Beijing’s tech hubs. Recently, Moonshot AI’s Kimi K3 model has garnered international attention for its proficiency in complex coding tasks and agentic workflows—areas previously dominated by American leaders such as OpenAI’s GPT-4 and Anthropic’s Claude 3.5 Sonnet.
Other Chinese giants, including Alibaba with its Qwen series, Baidu with Ernie Bot, and the startup DeepSeek, have also released models that frequently top global leaderboards in reasoning and mathematical benchmarks. This surge in performance has led to increased scrutiny from U.S. policymakers regarding the methods used to achieve such rapid progress. While Chinese developers attribute their success to architectural innovations and massive domestic datasets, some American firms and government officials suspect that "model distillation"—the process of using outputs from a superior model to train a smaller or competing one—is being used to bypass years of costly research and development.
A Chronology of Escalating Technology Restrictions
The threat of sanctions against AI developers is the latest chapter in a multi-year effort by the U.S. government to maintain its "chokepoint" over critical technologies. The trajectory of these restrictions illustrates a broadening scope of intervention:
- October 2022: The Biden administration introduced sweeping export controls, preventing Chinese firms from purchasing advanced logic chips and memory chips, most notably NVIDIA’s A100 and H100 GPUs, which are essential for training large language models (LLMs).
- August 2023: An executive order was signed to restrict U.S. venture capital and private equity investment into Chinese sectors involving semiconductors, quantum computing, and artificial intelligence.
- October 2023: The Department of Commerce further tightened export controls, closing loopholes that allowed the shipment of slightly downgraded "China-specific" chips.
- Late 2024: Reports surfaced that the U.S. was considering "guardrail" restrictions on the export of AI model weights themselves, fearing that open-source releases could be weaponized by adversarial nations.
- January 2025: Following the transition to the Trump administration, Treasury Secretary Scott Bessent’s comments suggest a more aggressive, punitive approach involving direct financial sanctions on specific corporate entities.
The move toward sanctions represents a shift from defensive measures (denying access to tools) to offensive measures (penalizing the use of the results).
The Distillation Debate: Intellectual Property or Common Practice?
At the heart of the dispute is the technical concept of "model distillation." In the AI industry, distillation typically involves a "teacher" model (a large, high-parameter system) providing responses that are then used to train a "student" model (a smaller, more efficient system). This allows the student model to mimic the reasoning and accuracy of the teacher while requiring significantly less computational power.
U.S. developers, including OpenAI, have updated their terms of service to explicitly prohibit the use of their model outputs to train competing AI systems. They argue that this practice constitutes a form of intellectual property theft, as it allows competitors to "harvest" the intelligence of a model that cost hundreds of millions of dollars to train.
However, this position is not universally accepted within the tech community. Microsoft CEO Satya Nadella has recently raised questions about the consistency of this stance. Nadella noted the irony of major AI companies relying on "fair use" principles to train their own models on vast quantities of public internet data—often without compensating the original content creators—while simultaneously attempting to impose highly restrictive terms on how their own outputs are used.
Furthermore, Clem Delangue, CEO of Hugging Face—the world’s largest platform for open-source AI—has argued that distillation is only a minor contributor to China’s AI advancements. According to Delangue, China’s progress is better explained by its massive pool of research talent, substantial state-led investment, and a cultural embrace of open-source collaboration that allows developers to iterate faster than their counterparts in more siloed environments.
Supporting Data: The Scale of the AI Competitive Landscape
The intensity of the U.S. response is underscored by the sheer scale of investment and the speed of development in the sector. According to data from the World Intellectual Property Organization (WIPO), China filed over 38,000 AI-related patents in 2023, significantly outpacing the United States in raw numbers, though the U.S. still leads in "high-impact" foundational research citations.
Market analysis suggests that the cost of training a state-of-the-art foundational model has skyrocketed from approximately $10 million in 2020 to upwards of $100 million for models like GPT-4, with estimates for next-generation models exceeding $1 billion. From the perspective of the U.S. Treasury, allowing foreign competitors to bypass these costs through distillation or data scraping represents not just a security risk, but a massive economic transfer of value.
Conversely, Chinese AI startups have seen a surge in private funding despite U.S. investment bans. Moonshot AI, for instance, raised over $1 billion in a 2024 funding round led by Alibaba and HongShan (formerly Sequoia China), valuing the company at roughly $2.5 million. The rapid valuation growth of these "AI Tigers" in China has heightened fears in Washington that American technological hegemony is under immediate threat.
Domestic Legal Complications and the Fair Use Paradox
The U.S. government’s push to protect AI intellectual property is complicated by a flurry of domestic legal challenges. While Washington warns Beijing against theft, American AI companies are themselves being accused of systematic copyright infringement at home.
In a landmark case, Anthropic recently received preliminary approval to begin payments under a $1.5 billion settlement with a group of authors. A federal judge found evidence that the company had illegally downloaded and stored thousands of copyrighted books to train its Claude models. Similarly, the New York Times is currently engaged in a high-stakes lawsuit against OpenAI and Microsoft, alleging that the developers used millions of the newspaper’s articles without permission.
These domestic cases create a diplomatic and legal "fair use" paradox. If U.S. courts decide that training an AI on copyrighted data is a protected "fair use," it becomes legally difficult for the U.S. Treasury to argue that Chinese companies using AI-generated outputs for training are committing "theft." If, however, the U.S. enforces strict IP protections for model outputs, it may inadvertently weaken the legal defense of its own companies against domestic copyright holders.
Potential Implications of Sanctions
If the Treasury Department proceeds with sanctions, the implications for the global AI ecosystem would be profound. Sanctions could take several forms, ranging from adding companies to the "Entity List" (which prevents them from doing business with U.S. firms) to more severe financial sanctions that freeze assets and block access to the U.S. dollar-denominated financial system.
- The "Splinternet" Effect: Sanctions could accelerate the decoupling of the global tech stack. China would likely double down on its "self-reliance" drive, creating a completely independent ecosystem of chips, software, and data standards.
- Impact on Research Collaboration: AI research has historically been a highly collaborative, international endeavor. Sanctions could lead to a "chilling effect" on academic exchanges, with U.S.-based researchers fearing legal repercussions for collaborating with Chinese peers.
- Retaliation Against U.S. Tech: Beijing has already demonstrated a willingness to retaliate, as seen in the restrictions placed on Micron Technology and the investigation into Intel. Further U.S. sanctions could lead to a ban on American AI services or cloud providers operating within the Chinese market.
- Operational Risks for Multinationals: Companies like Microsoft, which has a significant research presence in China (MSRA), could find themselves caught in a regulatory crossfire, forced to choose between their global R&D footprint and compliance with Treasury Department mandates.
Conclusion: A New Era of Economic Statecraft
Secretary Bessent’s comments reflect a growing consensus in Washington that artificial intelligence is the primary "dual-use" technology of the 21st century—essential for both economic prosperity and national security. By framing AI development through the lens of intellectual property theft and national sanctions, the Trump administration is signaling that it views the AI race as a zero-sum game.
The challenge for the Treasury Department will be defining the threshold for "theft" in an industry where the lines between inspiration, distillation, and infringement are technically blurry and legally untested. As the U.S. prepares to potentially blacklist some of China’s most promising tech startups, the global community watches to see if these measures will preserve American leadership or simply hasten the arrival of a bifurcated technological world. For now, the warning from the Treasury is clear: the software that powers the future is now a front line in the battle for global influence.
