Editor's Note: The AI competition is transitioning from a battle of model capabilities to a more complex systemic competition.
This article discusses Anthropic's latest assessment of the AI competition between China and the U.S. The author believes that the next two to three years will be a crucial window for shaping the frontier AI landscape: on one hand, the U.S. and its allies still hold the advantage in advanced chips, model capabilities, capital investment, and global tech stack; on the other hand, Chinese AI labs are continuously approaching the frontier with talent, data, engineering efficiency, and rapid catch-up capabilities.
Based on this, Anthropic believes the current core task is to maintain its leading edge in computing power and model capabilities. This includes both strengthening export controls on advanced chips and restricting technology spillover paths such as overseas data centers, chip transfers, and model distillation. Otherwise, Chinese AI companies may further narrow the gap with the U.S. frontier models by gaining access to computing power and replicating model capabilities by 2028.
The article also presents a broader industry judgment: as AI enters an acceleration phase of capabilities, the focus of competition is no longer just "who has the strongest model," but who can translate model capabilities into infrastructure, industrial efficiency, the global market, and governance rules. The closer AI technology gets to general-purpose capabilities, the more critical factors such as the chip supply chain, capital investment, policy tools, and global distribution networks behind it will become key variables determining the future landscape.
Below is the original article:
We have released a new paper outlining our views on the U.S.-China AI competition.
The United States and its allies need to maintain a relative edge over major competitors such as China in the AI field. As AI capabilities rapidly improve, this technology will soon deeply impact social governance, national security, and the international power structure. At the same time, the pace of AI development is accelerating, leaving little time for all parties to set competition rules, manage tech risks, and shape a global governance framework. It is in this context that we propose the measures necessary to ensure U.S. leadership.
One of the most critical elements in AI development is access to computing chips used to train models, known as "computing power." As the most advanced chips are mainly developed by companies within the U.S. and its allies, the U.S. government currently restricts China's access to these chips through export controls. Recent experiences have shown that these control measures have had a significant effect. In fact, Chinese AI labs have been able to develop models close to the U.S. level primarily by leveraging their talent advantage, exploiting export control loopholes, and engaging in large-scale model distillation -- by extracting outputs and capabilities from U.S. models to quickly replicate some technological achievements.
In this article, we present two scenarios for the world in 2028. We anticipate that by then, transformative AI systems will have emerged.
In the first scenario, the U.S. successfully maintains its computational power advantage. Policymakers further tighten export controls, reducing China's space to acquire U.S. cutting-edge capabilities through methods like model distillation, and accelerating the U.S. and its allies' adoption of AI. In this world, the U.S.-led tech ecosystem can more significantly influence AI's rules, standards, and governance frameworks. It is also in this scenario that the U.S. is more likely to engage in effective communication with China on AI security; to the extent feasible, we support this.
In the second scenario, the U.S. fails to take sufficient action. Policymakers fail to block China's avenues to advanced computing, and Chinese AI companies swiftly exploit these opportunities, catching up to the AI frontier, and even achieving leadership in some areas. In this world, rules and standards for AI will be contested by more countries, and the most advanced models may be used for broader-scale social governance, cyber actions, and security capabilities building. Even though this situation rests on U.S. computational power and technology spillover, it does not align with the long-term interests of the U.S. and its allies.
The U.S. and its allies enter the AI competition with a strong advantage. The key tools needed for AI primacy have been developed by a highly innovative enterprise ecosystem within the U.S. and its allies' system. Past success means that the most critical current task is largely about not squandering existing advantages: not making it easier for China to catch up.
Two Scenarios of U.S.-China AI Competition in 2028
Abstract
The development and deployment of AI will determine the future direction of global technological rules, industry standards, and governance frameworks. Whoever can maintain a lead in the AI field is more likely to shape the operation of these systems.
Currently, the U.S. and its allies have a significant lead in computational power. Computational power is one of the most critical elements for developing cutting-edge AI models. This lead stems from technological innovations in the U.S. and its allies and is supported by bipartisan export control policies in the U.S. But in terms of model intelligence level, Chinese AI labs are not far behind. We focus on China's AI development not to deny the capabilities and contributions of the Chinese people and the Chinese AI community but because China is the only country, besides the U.S., with ample resources and top talent that are systematically catching up with cutting-edge AI.
China has already applied AI technology in areas such as information censorship, social governance, cybersecurity, and military capacity building. Chinese AI labs have world-class talent. What truly limits their continued catch-up is computational constraints. The reason Chinese labs have been able to stay close to the frontier is partly because they have exploited loopholes in U.S. export control policies and have accelerated their model training and capability catch-up by distilling part of the capabilities from U.S. models on a large scale.
As computing power rapidly expands, AI is increasingly being used to enhance new model training, and we are entering a period of rapid acceleration in AI capabilities. The so-called "Genius Country in Data Centers" — the transformative AI intelligence level as we understand it — may be closer than ever. This acceleration makes policy action more urgent.
So far, due to ongoing issues such as export control evasion and model distillation, the Chinese AI system has continued to stay close to the cutting edge curve. However, if the United States and its allies take action now while addressing issues of accessing computing power and model capacity spillover, it is still possible to secure a 12 to 24-month lead in frontier capabilities. By 2028, such a lead will have significant strategic importance. This advantage will also enhance the ability of the United States and Chinese AI experts to communicate around AI security and governance, and we support this engagement. However, the window of opportunity to secure this lead will not always be open.
Here, we present two possible scenarios for the 2028 U.S.-China AI competition status. The first scenario is that the United States and its allies establish a significant lead in model intelligence, application adoption, and global deployment. If policymakers take action now, tighten control over advanced computing power in Chinese labs, reduce their space to catch up by distilling the best AI models from the United States, and accelerate the adoption of AI by the United States and its allies, this scenario may be realized.
The second scenario is that China is competitive in a near-frontier position. If policymakers do not continue to advance on the existing lead, or relax restrictions on Chinese companies accessing advanced computing power, this scenario will occur.
Many in the U.S. Congress and the Trump administration already support export controls, countering model distillation attacks, and promoting the overseas expansion of U.S. AI technology stack. With these policies advancing, we hope the United States and its allies will ensure a significant lead by 2028, avoiding a highly competitive race with China that will be neck and neck just two years later.
The Necessity of Maintaining Leadership
We expect that cutting-edge AI will have profound economic and social impacts in the coming years, as described in "Machines of Loving Grace" and "The Adolescence of Technology." Our mission is to ensure that humanity can safely and beneficially navigate the transition to transformative AI. We believe that a successful transition will bring significant breakthroughs in medicine, innovation, and economic growth.
Security and Governance Risks in AI Development
The success of this transition will partly depend on which ecosystem builds the most powerful systems first. The industry landscape, regulatory environment, and governance framework in which cutting-edge AI operates will shape the rules of engagement for the development and deployment of this technology. In turn, these rules will also influence the safety of the technology, who it secures, and whose interests it ultimately serves.
If the AI frontier is primarily set by systems that aim to leverage the technology for military advantage, cyber operations, social governance, and information control, then this technological transformation will face higher levels of uncertainty and security risks.
Historically, the ability for extensive governance and surveillance has often been limited by the costs of human enforcement. Powerful AI systems could lower such costs, enabling automated governance, identification, and decision-making at a larger scale. Therefore, China's leadership in AI could have a significant impact on the global AI governance and security landscape.
China holds vast economic, military, and national governance resources. It is also the only country besides the United States with well-resourced and highly concentrated talent AI labs that are catching up to the frontier. Additionally, China is actively positioning itself as a leading AI powerhouse. Beijing has already invested billions of dollars into China's AI and semiconductor industries.
China has already applied AI systems in areas such as information censorship, social governance, cyber operations, and security capability building. The deployment of relevant technologies in certain regions, including facial recognition, biometric data collection, and communication monitoring, also demonstrates the potential application of AI in large-scale governance. Cutting-edge AI systems will reduce the maintenance costs of these capabilities, expand their coverage, and increase automation. As these technologies spread overseas, AI may be used by more countries to enhance governance and surveillance capabilities. The AI frontier led by China could significantly alter the global technological usage and governance patterns.
AI is a dual-use technology
The frontier of AI will shape the future balance of military power. China views AI as a critical variable on the future battlefield and is advancing the intelligentization of its military systems. Strategic planners in the Chinese military see the "intelligentization" of military power as a crucial path to catch up and ultimately enhance their military capabilities. The Chinese military has begun procuring AI systems developed by Chinese companies for military purposes, including deploying the DeepSeek model to coordinate swarms of drones and enhance cyber operational capabilities.
These capabilities will not diffuse slowly. When a new model reaches a new capability level in autonomous targeting, vulnerability discovery, or cluster coordination, the side that masters it can operationalize it within weeks, not years.
Risks will be further compounded as cutting-edge AI becomes an accelerator for other key technologies. Advanced AI models will be able to compress the R&D cycles in semiconductor, biotech, and advanced materials. Leadership in cutting-edge AI will allow a nation to continually expand its advantage across the entire national security technology stack.
If a Chinese AI lab develops a model reaching the Claude Mythos Preview level before a U.S. lab, China would be the first to possess a system capable of autonomously discovering and chaining software vulnerabilities, potentially leveraging it to further enhance its cyber operational capabilities. The capabilities of future models will increase exponentially, therefore posing a larger impact on the security interests of the U.S. and other countries.
Coordinated Competition Could Weaken the Incentive for Responsible AI
A coordinated competition between U.S. and Chinese AI labs may make industry- and government-led security and governance efforts more challenging. If Chinese labs closely follow or are on par with U.S. models, both U.S. and Chinese private AI firms may feel more pressure to release new models and products faster without undergoing thorough security assessments, potentially impeding efforts to deploy responsible AI. Governments of various countries may also be reluctant to enact policies that encourage responsible AI development and deployment out of concerns of falling behind.
While an increasing number of researchers in Chinese AI labs and the policy community are starting to focus on AI security risks, this trend has not translated into security practices comparable to U.S. labs. As of last year, out of the top 13 AI labs in China, only 3 have released security assessment results, with none disclosing Chemical, Biological, Radiological, and Nuclear (CBRN) risk assessments. The AI Standardization and Innovation Center (CAISI) found that under a common jailbreak technique, the DeepSeek R1-0528 model responds to 94% of obviously malicious requests, compared to 8% in the U.S. reference model. This pattern continues in recent model releases. For example, an independent evaluation of Moonshot's Kimi K2.5 released in April this year found a higher proportion of CBRN-related requests accepted by this model compared to leading U.S. models.
More critically, Chinese labs frequently release models with dual-use capabilities in open weight form. Once the model's open weights are released, existing security safeguards could be stripped, allowing any state or non-state actor to exploit the model for malicious purposes, including cyberattacks and CBRN misuse, undermining the original safeguards designed to prevent such misuse.
Our Policy Objective: Create and Sustain a Leadership Advantage for the U.S. and Its Allies
We support the United States and other countries in adopting policies to establish and maintain a security and near-term lead relative to China in terms of intelligence, domestic adoption, and global distribution. This lead is crucial for protecting the national security interests of the United States and its allies, and for preventing the misuse of AI technology. Doing so is also a fundamental prerequisite to ensure that the United States and its allies can secure a favorable position in future global AI governance.
Anthropic deeply respects the Chinese people and acknowledges the achievements of the Chinese AI community. We hope for peaceful relations between China and the world. Our concerns specifically focus on the risks that any powerful national system may pose to global security and governance once it acquires cutting-edge AI systems.
Opportunities for AI Security Engagement
Where feasible, Anthropic supports international AI security dialogues with Chinese AI experts. The world has a common interest in secure AI regardless of where it is developed and deployed. Cutting-edge AI systems may bring about a range of risks that require communication between the United States and China. Identifying common challenges, advancing relevant concepts to prepare for and mitigate these risks, is in the mutual interest of both parties.
The prospects for constructive engagement are best when the United States maintains a significant capability advantage. Establishing a lead in a responsible manner in the development and deployment of cutting-edge AI will enhance the U.S.'s ability to influence AI security practices in China and other regions.
Alert from Mythos Preview
The Mythos Preview was released in April this year as part of Project Glasswing to select partners. It indicates that an era of accelerating capabilities has arrived, making policy action more urgent. After gaining access to the model, Firefox patched a greater number of security vulnerabilities last month than the total fixed in the year 2025, nearly 20 times the monthly average of security flaw fixes in 2025. In response to this model, a Chinese cybersecurity analyst wrote that China was "still sharpening the knife, and the other side suddenly set up a fully automatic Gatling machine gun."
The advancement in cutting-edge AI capabilities will rapidly approach the transformative AI vision of "genius-in-a-data-center." This acceleration will be driven by the logic of the law of accelerating returns: as computational power and data input increase, model performance will predictably improve, and AI itself will increasingly be used to expedite the development of new models.
We are likely to look back in the future and see 2026 as a window of opportunity for the United States to achieve breakthrough AI leadership. U.S. labs hold the most advanced AI models, have a significant lead in the advanced AI chip quantity and quality needed for the forefront, and with their income and financing, have a huge capital advantage to support relevant investments. Chinese labs do possess a real advantage: world-class innovative talent, abundant and low-cost energy, and vast amounts of data. All necessary conditions for advancing cutting-edge intelligence. However, they do not have enough domestic computational power to compete, nor do they have enough income and capital to fund this competition.
The Four Fronts of Competition
The United States and China are engaging in a strategic competition for advantage in cutting-edge technologies such as AI. Public statements from Beijing and Washington reflect this assessment. Referring to this competition as a "race" may give a misleading impression, as if there were a finish line that, once crossed, would allow one side to definitively secure victory. In reality, this will be an ongoing competition for advantage. Whether democratic or non-democratic nations will be more successful in shaping the values, rules, and norms of the AI era in the future will still depend on the course of this long-term competition.
This competition is unfolding on four fronts:
Capability: Which countries can develop the most powerful AI models.
Domestic Uptake: Which countries can most effectively integrate AI into their commercial and public sectors.
Global Deployment: Which countries can deploy an AI technology stack that supports the global economy.
Resilience: Which countries can maintain political stability during economic transformation.
Among these four fronts, capability is the most critical. We anticipate that the capabilities of cutting-edge models will have the most profound impact on geopolitical competition. Model capability is also a core factor driving market adoption and global deployment.
However, capability alone is not sufficient. If China can more quickly and effectively integrate near-state-of-the-art AI systems into its economy and security apparatus, drive the global adoption of low-cost, subsidized AI, and prioritize "embodied intelligence," China may gain an advantage significant enough to offset any model intelligence gap. Beijing's "AI+" initiative and its focus on "embodied intelligence" reflect its strong emphasis on incorporating cutting-edge intelligence into the economic and national systems. The Trump administration's AI initiatives, particularly the emphasis on "advancing the export of American AI technology," also underscore the strategic advantage of promoting global adoption.
While this article does not focus on the "resilience" front, we believe it will become a significant aspect of the AI competition. Maintaining stability, cohesion, and effective policymaking during this period will be a key advantage. Conversely, countries that fail to achieve this will face a vulnerability.
Current Competitive Landscape
Compute power—meaning the advanced semiconductors needed to train and deploy cutting-edge AI—is a key input on each of the above competition fronts. The global competition for AI leadership is largely a competition for compute power. Over the past decade, model capabilities have been enhanced as compute scales have increased; historically, much of the performance gains in AI capabilities have come from using compute at a larger scale.
In addition, computing power is used not only for training new models but also to support user AI inference capabilities. Whether it is training the smartest models or deploying these models in the commercial and national security sectors, computing power is crucial. Top talent, vast amounts of data, and breakthrough algorithms are all vital for the AI race; however, without sufficient computing power, these investments are challenging to realize.
Currently, democratic nations are winning the battle for computing power leadership. Some are concerned that export controls may accelerate China's efforts to develop a local advanced chip supply chain, but there is little evidence to suggest that China's self-sufficiency efforts can challenge the United States and its allies' leading position in advanced computing technology. Even before the implementation of export controls, Beijing had invested heavily in the Chinese semiconductor industry and launched significant industrial policies such as "Made in China 2025" and the National Integrated Circuit Industry Investment Fund. Despite these state-backed investments, Chinese AI labs and chip manufacturers are still constrained by the U.S. and its allies' export controls on advanced chips and semiconductor manufacturing equipment.
The result is that the computing power gap appears to be widening. An analysis of Huawei and NVIDIA product roadmaps found that in terms of total processing performance, by 2026, Huawei can only produce products equivalent to 4% of NVIDIA's total computing power, decreasing to 2% by 2027. More importantly, NVIDIA is just a part of the U.S. and its allies' computing ecosystem. Google and Amazon are also ramping up production of their own chips, namely TPU and Trainium, to meet the needs of U.S. frontline AI labs and their customers.
Further exacerbating China's computing power shortage is the limited progress China has made in several of the most complex aspects of the semiconductor supply chain. Without access to Extreme Ultraviolet Lithography (EUV) technology, especially as policymakers further block off Deep Ultraviolet Lithography (DUV) technology and its service and maintenance vulnerabilities, Chinese chip manufacturers will struggle to produce an adequate quantity and quality of chips to challenge the U.S. computing dominance. China's inability to mass-produce high-bandwidth memory further widens this gap. One study estimates that if the U.S. strengthens restrictions on China's access to U.S. computing capabilities, the computing power available to the U.S. will be approximately 11 times that of China's AI industry.
How Democratic Nations Can Establish Leadership: Business Innovation and Effective Public Policy
Computing power leadership primarily stems from two reasons.
The first reason is the continuous innovation of companies such as NVIDIA, AMD, Micron, TSMC, Samsung, ASML, and others in democratic economies like the United States, Japan, South Korea, Taiwan, and the Netherlands. It is these companies that collectively have built the unique technologies required for the world's most advanced semiconductors. Without these engineering breakthroughs and decades of sustained R&D investment, today's AI achievements would not be possible.
The second reason is that the past three U.S. administrations have taken forward-looking and decisive policy actions. Policy actions driven by both parties have restricted Chinese-controlled companies from accessing the U.S. AI technology stack, safeguarding the innovation engine of the United States and its allies. Our CEO has also publicly commented on the importance of export controls. In recent years, these controls have limited the sale of cutting-edge AI chips and semiconductor manufacturing equipment to China, even though Beijing has invested significant national resources in this field, the advancement of China's cutting-edge AI is still constrained. Without action to restrict China's access to U.S. computing power, China might have had all the conditions to develop AI on par with or even stronger than that of the United States.
Some observers are concerned that restricting access to computing power will force Chinese AI laboratories to innovate in other directions, thereby weakening the U.S. lead. Chinese laboratories are indeed innovating, but so far, these innovations have not been sufficient to bridge their computing power gap. Algorithmic improvements are both a function of computing power and a multiplier of computing power, not a substitute for it. The discovery of these algorithmic advancements itself is a process highly dependent on computing power: more computing power means laboratories can run more experiments to discover more algorithmic improvements. As more cutting-edge models are involved in AI research and development, this cycle will tighten further, with cutting-edge models helping build their next generation. In short, computing power advantage will further translate into algorithmic advantage, and ultimately into a sustained lead in AI itself.
Currently, U.S. cutting-edge systems are estimated to be at least several months ahead of China's top models in terms of intelligence level, although such estimates inevitably involve uncertainty. While China's open-source models have received much attention, they still lag in enterprise adoption compared to closed-source cutting-edge models, and public-market investors have started to pay attention to their commercialization issues. Furthermore, Chinese AI laboratories seem to be moving away from the open-source path, opting instead to keep the best models private.
Chinese AI leaders have also acknowledged the impact of export controls and the critical need for U.S. chips. Executives from top Chinese AI laboratories have expressed concerns that China will fall further behind due to computing power restrictions. Senior executives at leading AI laboratories in China have identified computing power scarcity as a primary constraint to accelerating model capabilities and consider export controls as the cause of this constraint. A senior executive from a Chinese major cloud provider stated that supplying U.S. chips subject to export controls to China would have a "huge, really huge" impact, adding that any supply gap would severely affect China's AI development; he also refuted concerns that "importing U.S. chips would slow down China's efforts toward autonomy." The main voices advocating that "export controls are ineffective" in China seem to come more from official pronouncements and state-owned media, possibly aimed at influencing U.S. policy makers.
How China Maintains Competitiveness: Policy Gaps Still Exist
While export controls have been effective in shaping the current landscape, their impact remains insufficient. Despite China's inability to domestically manufacture enough advanced chips and the inability to legally purchase these chips abroad, Chinese AI labs still maintain a close position to the cutting edge in model intelligence through two alternative means.
The first method is evasive compute acquisition, including smuggling AI chips directly into China or accessing overseas data centers. The second method is illicit model access, namely conducting distillation attacks on U.S. cutting-edge models and using these models as tools to expedite their own AI development.
It is an open secret that China circumvents U.S. export controls. For example, U.S. federal prosecutors have accused a Supermicro co-founder and two others of transferring $2.5 billion worth of servers containing advanced U.S. chips to China. According to U.S. government and media reports, DeepSeek trained its latest model using advanced U.S. chips banned from sale to China. The Financial Times reported that Alibaba and ByteDance are now training their flagship models in data centers in Southeast Asia using restricted U.S. chips. The current controls do not cover this pathway because U.S. export laws mainly regulate chip sales rather than remote chip access. The U.S. export control system is struggling to address the issue of Chinese AI labs gaining access to U.S. advanced computing power.
Distillation attacks are another method used to catch up with U.S. counterparts and mitigate the impact of export controls. In this practice, Chinese labs create numerous fake accounts to bypass access controls on U.S. AI models and systematically collect the outputs of these models to replicate cutting-edge capabilities. This approach allows relevant labs to hitch a ride on decades of U.S. foundational research, tens of billions of dollars in investment, and cutting-edge model outputs developed collaboratively by top engineers worldwide. As a result, China can obtain close-to-cutting-edge capabilities at a very low cost, a cost that is essentially subsidized by the U.S. From a long-term national security perspective, this is akin to systematic industrial espionage on critical technologies. OpenAI, Google, Anthropic, and the Frontier Model Forum have all publicly condemned distillation attack practices.
Chinese AI experts have also openly acknowledged the scale and significance of distillation attacks on China's AI development. A recent article in a state-owned media described distillation attacks on U.S. models as a "backdoor" that Chinese AI labs rely on and referred to it as a core part of their business model. A former ByteDance researcher stated that Chinese AI labs use distillation as a shortcut to training models, thereby avoiding the investment in building their own data pipelines.
U.S. policymakers have swiftly acted to address this threat. The White House Office of Science and Technology Policy issued a memorandum on distillation attacks. The White House, the U.S. Department of Defense, and senior members of Congress have also expressed concern about this issue. Recently, legislation proposed by the U.S. House Foreign Affairs Committee aims to address distillation attacks and has received unanimous approval in the committee.
If U.S. and allied policymakers can block the two channels supporting China's AI model development—evasive compute supply and illicit model access—we may have a rare opportunity to gain a decisive advantage that locks in leadership.
Two Scenarios for 2028
Below, we describe two hypothetical future scenarios to illustrate how today's policy actions will shape the competitive landscape in 2028.
Scenario One: U.S. and Allies Hold Overwhelming and Expanding Leadership
The U.S.'s compute advantage remains robust. Despite China's increased state support for the semiconductor industry, Chinese chipmakers still lag years behind the U.S. and its allies, in part because they cannot access advanced semiconductor manufacturing equipment, related services, and maintenance. As the U.S. and its allies bring chip manufacturing capabilities online and advanced chipmakers continue to develop more efficient, higher-performance chips, the U.S.-China compute gap is widening.
At the same time, U.S. policymakers take action to close loopholes in U.S. economic security tools. With more enforcement resources, efforts to smuggle chips into China and to access export-controlled chips in overseas data centers are becoming increasingly challenging.
Therefore, U.S. AI models are leading by 12 to 24 months in intelligence capabilities, and the leadership gap is widening. A few AI labs are at the forefront with the smartest, most powerful, and performance-optimized models, all located in the U.S. The "genius country in data centers" has become a reality in key industries such as cybersecurity, finance, healthcare, and life sciences.
When U.S. leading labs release new models in 2028 with dramatically improved capabilities—a relative impact akin to the April 2026 Mythos Preview—China may not achieve similar AI capabilities until 2029 or 2030. This will provide democratic nations with critical buffer time to set rules and norms for cutting-edge AI systems.
U.S. AI becomes the foundation of the global economy, driving new economic and scientific vitality. The Trump administration's efforts to promote domestic AI adoption and U.S. AI exports have paid off, with robust AI widely adopted domestically and internationally, leading to unprecedented economic growth and technological advancement. Global adoption of U.S. AI has significantly increased. The democratic nations' lead in capabilities and compute power means that Chinese AI firms struggle to compete globally outside of a few national markets. The world's top cutting-edge AI systems shaped by democratic values also make it harder for some countries to exploit AI systems to violate rights and civil liberties.
The cybersecurity and broader national security advantage further expands. Cybersecurity professionals in the public and private sectors use advanced AI systems to shrink the attack surface of the United States and other democratic nations, weakening China's ability to gain and maintain a foothold in relevant systems, thus making national security assets, intellectual property, and communication networks more secure. The overwhelming AI advantage of the United States also becomes a key force in mitigating external risks.
A self-reinforcing cycle will further entrench the leadership of democratic nations. The overwhelming AI advantage makes the United States and its allies more attractive partners. This alliance expands the market for American AI and also broadens the coalition that establishes global AI norms. In turn, this promotes the development and deployment of AI systems that are secure, reliable, and protective of civil liberties. The world's top technological and scientific talent continues to flow to centers of advanced technology. As a result, the United States gains crucial leverage to drive cooperation with Beijing on key issues such as AI governance, strategic competition, and trade.
This cycle will continue to self-reinforce: leadership advantage strengthens the alliance, and the alliance further enhances the leadership advantage; the international order led by democratic nations will also be anchored in the transition to transformative AI.
Scenario 2: AI Ecosystem Under Chinese Control Catches Up with the United States
The AI developed and deployed by China is nearing the frontier in model intelligence. Despite weaker semiconductor production capacity, models trained in Chinese AI labs lag only a few months behind American models. Ongoing distillation attacks, overseas compute access, enforcement gaps in semiconductor manufacturing equipment exports, and relaxation of U.S. controls on semiconductor exports have all helped China in its catch-up efforts. Continued access to cutting-edge American AI for AI research has also allowed Chinese AI labs to narrow the gap and approach their American counterparts.
The adoption at both commercial and national levels is rapidly advancing. Beijing is driving nationwide domestic adoption through its "AI+" policy. Even though China's AI model capabilities are slightly behind those of the United States, China's efforts to drive adoption have already borne fruit. Consequently, China is able to more advantageously deploy near-frontier AI capabilities in the economic, military, and technological fields, shifting the balance of power towards China.
China's AI-enabled cyber capabilities pose a significant threat. China integrates AI-enabled cyber capabilities into its already highly mature cyber power structure, enabling the Chinese military to remain a threatening cyber competitor. These cyber-enabled activities have gained more access to critical infrastructure in the United States and most countries worldwide, allowing them to disrupt key national security and societal functions. As AI becomes more deeply integrated into the most critical systems, even though democratic nations were the first to develop this technology, they cannot gain an advantage over China in AI security.
Beijing is winning global adoption with its cost and localization deployment flexibility. Huawei's and Alibaba's data centers are widely present globally, especially in low-cost markets in the Global South, but not limited to these regions. These data centers scale capacity based on older chips, and China's ability to export these chips comes from its domestic market being able to acquire US chips with export licenses, smuggle chips into China, or remotely access overseas data centers to meet demand. These data centers host second-tier models produced by Chinese labs, which are not top-tier but offer lower prices and are still effective.
Similar to Huawei's past "cheap and cheerful" approach, China's close-to-the-edge models and hardware support a significant and rapidly growing part of the global economy. This infrastructure advantage will allow China to wield significant influence in relevant markets.
How to Stay Ahead
To ensure ending up in the first scenario, we support the following policy action directions.
Plug the leaks: chip smuggling, overseas data center access, and semiconductor manufacturing equipment.
Currently, Chinese labs acquire controlled US chips through smuggling and overseas data center access, while the gap in semiconductor manufacturing equipment control is accelerating its efforts towards autonomy. Tightening controls and increasing enforcement budgets would help plug these loopholes that support China's AI ecosystem. This will lower China's AI compute ceiling and consequently slow down its AI progress, thus maintaining and expanding the AI leadership advantage of democratic nations. It's worth noting that a lower compute ceiling may substantially weaken distillation attacks, as Chinese AI labs still need to reach a certain computing power threshold to effectively conduct illicit distillation.
Protect our innovation: restrict model access, curb distillation attacks.
US policymakers in Congress and the administration can continue to support relevant policy actions to penalize and curb distillation attacks from Chinese labs and take measures to help US labs themselves detect and stop distillation attacks. These measures can include legislatively clarifying distillation attacks as illegal acts and promoting threat intelligence and technology sharing among US peer labs and between labs and the US government. Curbing such behavior can substantially prolong the leadership advantage of democratic nations in the coming months and years.
Drive US AI exports.
As AI is increasingly adopted by global public and commercial sectors, the Trump administration should continue to push for trusted AI hardware and models developed and shaped by democratic principles to be adopted globally. Locking in trusted US infrastructure now can prevent the Chinese AI ecosystem from gaining the global foothold it needs for its competitive cost and adoption in the future.
Conclusion
The United States and its allies have developed the world's most powerful cutting-edge AI models and possess the most advanced AI key inputs globally. This has provided a significant advantage. If we can maintain priority access to these technologies, this advantage can continue to grow. However, if these technologies are directly transferred to competitors, this advantage will be lost. The decisions policymakers make this year will determine the future of transformative AI. We support those working to ensure that the United States and its democratic allies remain at the forefront in 2028.
Welcome to join the official BlockBeats community:
Telegram Subscription Group: https://t.me/theblockbeats
Telegram Discussion Group: https://t.me/BlockBeats_App
Official Twitter Account: https://twitter.com/BlockBeatsAsia
