Tightened Export Controls Could Keep US Ahead in AI by 2028

The competition in artificial intelligence is rapidly narrowing, with AI labs in China building models increasingly comparable to those in America, not through independent innovation, but by exploiting loopholes in existing export controls and conducting distillation attacks that extract innovations from American companies. A new paper from Anthropic focuses on a critical window of time, 2028, when transformative AI systems are expected to arrive, making the current moment decisive for establishing global leadership. “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 paper warns, highlighting the stakes of this technological race. Access to computer chips, or “compute,” is identified as the most important ingredient for AI development, and the US currently maintains an advantage due to its control over the most capable chips and the enforcement of export restrictions.

US Compute Advantage & Export Controls

The United States currently maintains a decisive advantage in the development of artificial intelligence, not through algorithmic innovation alone, but through control of the specialized computer chips, or “compute”, required to train increasingly complex models. While AI labs in the People’s Republic of China possess talent, their progress has been significantly constrained by limitations in accessing this essential hardware. Recent history demonstrates the success of US government export controls, though Chinese labs have attempted to circumvent these restrictions through exploiting loopholes and employing large-scale distillation attacks. These distillation attacks represent a particularly concerning trend; they illicitly extract innovations from American AI models to mimic their capabilities, effectively lessening the impact of compute restrictions. According to Anthropic, “AI 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.” This underscores the need for proactive policy measures to not only limit chip exports but also to disrupt these innovation-harvesting techniques.

Maintaining a lead in compute is paramount, as it directly impacts the pace of AI development and the potential for establishing favorable global norms. “The most important ingredient for developing AI is access to the computer chips on which the models are trained,” the report asserts, emphasizing that current US policy is “largely to avoid squandering our advantage: to decide not to make it easier for the CCP to catch up.” A substantial lead, estimated at a six-month lead in frontier capabilities by 2028, would not only accelerate domestic innovation but also facilitate constructive engagement with Chinese AI experts on safety and governance. However, the window of opportunity to secure this advantage is rapidly closing, demanding immediate and decisive action from policymakers to safeguard American leadership in this critical technological domain.

China’s AI Progress via Distillation Attacks

Artificial intelligence development currently sees a clear, though increasingly contested, lead held by the United States and its allies, primarily due to advantages in foundational compute power. These attacks represent a significant method for extracting innovation from US companies, allowing Chinese labs to rapidly close the gap in model intelligence. The urgency of this situation is underscored by projections for 2028, when transformative AI systems are anticipated to arrive. This timeframe creates a critical window for establishing global leadership, as the political and ethical implications of advanced AI become increasingly pronounced. The report highlights that China’s progress isn’t insurmountable, and continued vigilance is required to prevent China from catching up. The analysis presents two potential scenarios for 2028. In the first, proactive US policy, tightened export controls, disruption of distillation attacks, and accelerated domestic AI adoption, maintains a six-month lead for democracies.

This scenario would allow for constructive engagement with China on AI safety. Conversely, inaction could result in China achieving parity, shaping AI norms according to authoritarian principles and enabling “automated repression at scale.” “If the frontier is set by regimes that treat AI as an instrument of repression, military advantage over democracies, and domestic control, the transition is less likely to go well, for those regimes’ own citizens or anyone else.” The report concludes that a proactive approach is essential to ensure AI benefits humanity, rather than reinforcing authoritarian control.

Two Scenarios for 2028 AI Leadership

Anthropic, the AI safety and research company, is focusing attention on a pivotal moment in the development of artificial intelligence: 2028. The central argument revolves around maintaining a competitive edge over the People’s Republic of China and preventing the development of AI systems that could be used for widespread repression. The most critical factor, according to Anthropic, is access to “compute,” the computer chips necessary to train advanced AI models.

Currently, the US holds an advantage because the most capable chips are developed by American companies and are subject to export controls. “In this world, democracies set the rules and norms around AI,” and conversely, the second scenario envisions a failure to act, allowing Chinese AI firms to catch up and potentially surpass American capabilities. “In this world, AI norms and rules are shaped by authoritarian regimes, and the best models enable automated repression at scale.” Anthropic emphasizes that this outcome would not diminish the role of American compute, but rather see it utilized to bolster authoritarian control.

The paper highlights the urgency of the situation, noting that a substantial lead, approximately six months, in frontier AI capabilities by 2028 would be enormously advantageous. “Our mission is to ensure that humanity navigates the transition to transformative AI safely and beneficially,” and the company believes that democratic leadership is essential to achieving this goal. The stakes, Anthropic argues, extend beyond economic competition to encompass fundamental questions of security and freedom.

Authoritarian AI & Repressive Capabilities

The escalating competition between the United States and China in artificial intelligence carries profound implications for global security and civil liberties, extending far beyond economic dominance. A new paper from Anthropic details how advancements in AI could soon empower authoritarian governments with unprecedented tools for repression, shifting the balance of power internationally. The core concern isn’t simply that China is pursuing AI, but that its political system is uniquely positioned to exploit the technology for control. While Chinese AI labs possess considerable talent, their progress has largely depended on circumventing US export controls and employing large-scale distillation attacks to replicate American innovations. These attacks illicitly extract capabilities from US models, allowing China to remain competitive despite restrictions. This dynamic underscores a troubling reality: American innovation is, inadvertently, fueling the potential for authoritarian control elsewhere.

The paper outlines two scenarios: one where the US successfully maintains its compute advantage through stricter controls and disruption of these attacks, and another where China catches up, shaping AI norms according to its authoritarian principles. The report emphasizes that the CCP views AI as a means to strengthen its military and achieve dominance.

Accelerating AI Development & Frontier Intelligence

The prevailing narrative surrounding artificial intelligence often focuses on its potential for advancement, yet a recent analysis reveals a far more competitive and potentially fraught reality. This is accomplished through “large-scale distillation attacks that illicitly extract the innovations of American companies,” a process that effectively mimics US capabilities without the initial research and development investment. The paper centers on a critical juncture, 2028, when transformative AI systems are projected to arrive, establishing a decisive period for global leadership. Policymakers have, “by allowing export control evasions and distillation attacks,” permitted China’s AI progress to remain remarkably close to the American frontier. A proactive approach, tightening export controls and disrupting these extraction methods, could potentially secure “a six-month lead in frontier capabilities” by 2028, a margin considered enormously advantageous. The stakes extend beyond mere technological dominance.

Anthropic warns that if China were to lead in AI, it could usher in an era of “automated repression at scale,” leveraging the technology for domestic control and potentially exporting these capabilities to other authoritarian regimes. “Historically, the reach of authoritarian rule has been limited by its dependence on human enforcers,” the report notes, but powerful AI systems could remove this constraint.

Democracies’ Role in AI Norms & Governance

The accelerating pace of artificial intelligence development necessitates a proactive stance from democratic nations to shape global norms, a new paper from Anthropic argues, with the potential for widespread repression looming if authoritarian regimes gain dominance. Specifically, the paper suggests a six-month lead in frontier capabilities by 2028 would be enormously advantageous. The stakes are high, as the political system controlling the most advanced AI will dictate its development and deployment, influencing whether the technology prioritizes safety, security, and democratic values. “The CCP holds enormous power and influence at the helm of China’s economy, military, and the largest authoritarian state structure on Earth,” the report states, underscoring the urgency of safeguarding against a future where AI reinforces authoritarian control globally.