Chinese AI labs are rapidly closing the gap with American companies, not through superior hardware, but by leveraging talent and exploiting loopholes to extract innovations despite US export controls on computer chips. A new paper outlines two potential scenarios for global AI leadership, anticipating that transformative AI systems will arrive, creating a critical window for action. These actions are undermining export controls designed to limit China’s access to the advanced computing power needed to train leading-edge models. The analysis warns that failure to address this circumvention could lead to a future where AI norms are shaped by authoritarian regimes and the most powerful models enable repression at scale, even built on American compute. Democracies must lead in AI development and deployment, as these countries and political systems can shape the rules and norms that govern these systems.
US Compute Advantage Drives AI Leadership
America’s dominance in computer chip technology remains a key factor in maintaining its lead in artificial intelligence, despite increasingly sophisticated attempts by Chinese labs to close the gap. A new paper details how, while AI labs in the People’s Republic of China are rapidly developing their capabilities, they currently rely on circumventing US export controls and employing innovative methods to mimic American advancements. These efforts highlight a dynamic: the restrictions intended to limit China’s AI progress are being undermined by efforts to harvest innovations from US models. This urgency stems from the understanding that AI’s potential for both benefit and harm is accelerating.
Policymakers are focused on preventing a scenario where authoritarian regimes, specifically the Chinese Communist Party (CCP), dictate the norms and rules governing AI development. The paper asserts that the US and its allies must stay ahead of authoritarian governments like the Chinese Communist Party, warning of the potential for recent history suggests these controls have been successful. The competition isn’t simply about hardware; it’s about maintaining a substantial lead in compute, which refers to access to the processing power needed to train advanced AI models. The paper outlines two potential scenarios. In the first, the US tightens export controls, disrupts these attacks, and accelerates AI adoption within democratic nations, establishing a commanding lead. In this scenario, democracies set the rules and norms around AI. Conversely, if the US fails to act, Chinese labs could catch up and even surpass American capabilities. Ultimately, the paper argues that democracies, not authoritarian regimes, must lead in AI development and deployment, shaping the rules and norms that govern these systems, rather than becoming an instrument of repression, military advantage over democracies, and domestic control.
China’s AI Progress via Loopholes & Distillation
China’s pursuit of artificial intelligence leadership is increasingly defined by a strategy of circumventing restrictions and rapidly assimilating innovation, rather than solely relying on independent development. While US government export controls on advanced computer chips have been successful in limiting direct access to hardware, Chinese AI laboratories are actively employing methods to extract knowledge from American models. This process allows them to build comparable AI systems despite the hardware limitations, effectively mimicking capabilities without directly violating export regulations. These attacks represent a significant challenge to policymakers attempting to maintain a technological advantage. The competition is driven by the anticipated arrival of transformative AI within a currently unspecified timeframe, creating a limited window to establish dominance.
This isn’t simply a matter of technological prowess; the paper highlights the geopolitical implications, warning that AI will soon become powerful enough to be used to repress citizens at an unprecedented scale, and even to alter the balance of power among nations. The concern centers on the potential for authoritarian regimes, particularly the Chinese Communist Party, to leverage AI for domestic control and international influence. The authors are releasing a new paper that explains their views on the competition between the US and China, indicating a focused effort to address these challenges. Labs in China have managed to remain competitive by exploiting loopholes, utilizing distillation attacks, and leveraging talent, effectively harvesting innovations from US models. This allows them to build models close in intelligence to America’s despite the compute constraints.
The paper suggests that a substantial lead, a one-month lead in frontier capabilities, could be secured if the US and its allies act decisively to tighten controls and disrupt these attacks. However, the window of opportunity is closing rapidly, as the supply of compute expands and AI increasingly augments its own development. With the supply of compute expanding rapidly, and with AI being used increasingly to augment the training of new AI models, we’re entering a period of great acceleration in AI capabilities. The potential for a CCP-led AI frontier is viewed as a major threat, as it could enable automated repression on a far greater scale and strengthen authoritarian control globally.
That lead exists thanks to American and allied innovation, and to bipartisan US export controls that defend those innovations.
Two Scenarios for US-China AI Competition in
Researchers at a US policy organization are outlining potential futures for artificial intelligence competition between the United States and China, focusing on a critical window of opportunity to establish leadership. This success is being undermined by what the organization terms attacks, where Chinese AI labs illicitly extract innovations from American companies. These attacks allow Chinese labs to build models nearly as intelligent as those in the US, despite the restrictions on chip imports. In the first scenario, America has successfully defended its compute advantage. Policymakers have acted to tighten export controls further, disrupt China’s distillation attacks, and accelerate democracies’ adoption of AI. In the second scenario, America has chosen not to act. Policymakers have not tightened loopholes on the CCP’s access to compute, and AI firms in China have quickly taken advantage, catching up to the frontier and even overtaking America. Conversely, the second scenario envisions a failure to act, allowing China to leverage loopholes and rapidly advance its AI capabilities.
Labs in China have remained close by exploiting loopholes in US export control policies, and by carrying out large-scale distillation attacks that harvest the innovations of US models in order to mimic their capabilities.
Authoritarian AI: Risks of CCP Dominance
This circumvention isn’t simply about bypassing regulations; it highlights a critical vulnerability in the current approach to maintaining a technological advantage. The implications extend far beyond economic competition. The authors focus on the CCP as it is the regime most able to use frontier AI to cement authoritarianism; they do not seek to undermine the interests or ingenuity of the Chinese people. Already, the CCP is employing AI for censorship, surveillance, and cyberattacks. The stakes are high, and the window to secure a leading position in AI, and prevent its misuse, is rapidly closing.
We focus on the CCP as it is the regime that is most able to use frontier AI to cement authoritarianism; we do not seek to undermine the interests or ingenuity of the Chinese people.
Policy Tools: Export Controls & AI Adoption
Recent history suggests these controls have been successful. In fact, AI labs in China have only built models close in intelligence to America’s because of their talent, their ability to exploit loopholes around these export controls, and their large-scale distillation attacks that illicitly extract the innovations of American companies. The authors emphasize that democracies, not authoritarian regimes, must lead in AI development and deployment, shaping the rules and norms that govern these systems. Conversely, inaction, allowing loopholes and distillation attacks to continue, could lead to China catching up and potentially overtaking the US. Ultimately, the paper argues that avoiding a destabilizing neck-and-neck race with the CCP requires proactive policy decisions to safeguard America’s advantage and ensure a beneficial transition to transformative AI.
In fact, AI labs in China have only built models close in intelligence to America’s because of their talent, their knack for exploiting loopholes around these export controls, and their large-scale distillation attacks that illicitly extract the innovations of American companies.
Accelerating AI Capabilities & Urgent Policy Action
While restrictions on hardware shipments have demonstrably been successful, Chinese AI laboratories are circumventing these limitations by exploiting loopholes and utilizing distillation attacks that illicitly extract the innovations of American models. This timeframe is not merely academic; democracies, not authoritarian regimes, must lead in AI development and deployment, shaping the rules and norms that govern these systems. However, this lead is being eroded by China’s talent pool and its ability to exploit loopholes and employ distillation attacks. Many in Congress and the Trump administration have championed export controls, curbing distillation attacks, and exporting American AI, suggesting a path towards maintaining a one-month lead by 2030. A lead of this magnitude, the authors argue, would not only bolster US influence but also facilitate constructive engagement with China on AI safety and governance. The stakes are high, as the CCP’s ambition to become the leading AI power is fueled by substantial investment and a clear vision of.
AI will soon become powerful enough to be used to repress citizens at unprecedented scale, and even to alter the balance of power among nations .
