Sharper: China's AI Rise

Artificial intelligence has become a strategic priority for Beijing and Washington, yet both countries continue to pursue dramatically different strategies in their bid for global leadership. Ahead of Xi Jinping’s visit to Washington, major strategic decisions on the future of AI competition between the two countries loom, whether it is decisions over how—or whether—to restrict China’s access to cutting-edge chips and chipmaking technology or how the U.S. should regulate increasingly capable Chinese open-source models. In this edition of Sharper, CNAS experts weigh in on the major strategic questions, the practical implications, and how U.S. policy should be oriented toward China’s AI advancements.

Features:

Report | Superpowers and AGI

While immense focus has been paid to the possible achievement of AGI as AI systems grow increasingly capable, how AGI would affect geopolitical competition between the United States and China has received comparatively less rigorous attention. In Superpowers and AGI: How U.S.-China Competition Could Be Shaped by the Emergence of Artificial General Intelligence, Jacob Stokes considers how this competition would play out and the choices and risks both sides must contend with as the advent of AGI approaches.

Virtual Event | The AI Open-Weight Debate: Balancing Innovation and Security

While immense focus has been paid to the possible achievement of AGI as AI systems grow increasingly capable, how AGI would affect geopolitical competition between the United States and China has received comparatively less rigorous attention. In Superpowers and AGI: How U.S.-China Competition Could Be Shaped by the Emergence of Artificial General Intelligence, Jacob Stokes considers how this competition would play out and the choices and risks both sides must contend with as the advent of AGI approaches.

Report | Red Lines

No matter the relative advantage the American AI ecosystem may have over China, the global proliferation of frontier Chinese systems and their direct instrumentalization by the CCP pose a significant and growing threat to U.S. national security. In Red Lines: Understanding the National Security Risks of China’s Advanced AI, author Daniel Remler outlines a framework for measuring the absolute risks of AI and provides a detailed assessment of the capabilities, design choices, and security vulnerabilities of China’s most advanced systems across it. In the report, he argues that only by shifting away from a relative framing of an “AI race” can policymakers craft an effective response.

Report | Adversarial Distillation

Export controls have put China at a deficit in compute, complicating the CCP’s ability to compete on equal terms with the U.S. in AI development. One way Chinese developers have sought to get around these restrictions is adversarial distillation—gaining unauthorized access to American models and extracting their capabilities to make gains at an otherwise impossible speed and scale. In Adversarial Distillation: China’s Campaign to Extract American AI Capabilities, authors Daniel Remler and Ben Hayum analyze how adversarial distillation campaigns unfold and what steps the U.S. policy community can take to detect and deter them along the AI supply chain.

CNAS Insights:

MATCHing Policy to Strategy

If the United States wants lasting leverage, it should look upstream: to the chip manufacturing equipment. While controlling chip access restricts what China can acquire today, Janet Egan and Michelle Nie argue that curtailing semiconductor production will limit what it can produce for years to come. Both steps are vital to maintain and extend the U.S. and allied lead in AI.

Unpacking the H200 Export Policy

For years, U.S. export controls have restricted China’s access to advanced chips, forcing Chinese AI labs to train models on less capable hardware. Earlier this year, Janet Egan and James Sanders wrote that the decision to allow the sale of H200s and other advanced chips threatens a key advantage that currently distinguishes American AI capabilities from China’s.

Trump Should Talk to Xi About Military AI

Jacob Stokes and Daniel Remler argue that by talking about military AI, the United States can signal strength toward China and simultaneously prop up peace in a moment when it faces many tests. As autonomous systems grow increasingly capable, the authors write that Beijing and Washington could agree to measures designed to prevent accidents and enhance general stability without putting unrealistic constraints on military-technological innovation.

Commentary:

Analysis from Daniel Remler, Ruby Scanlon, Janet Egan, and Michelle Nie

In the News:

Insights from Vivek Chilukuri, Daniel Remler, and Ruby Scanlon