Finding common ground in the U.S.-China AI rivalry

概览

As Presidents Trump and Xi Jinping prepare to meet in Washington, artificial intelligence is high on the agenda. Both the United States and China face shared risks—from job losses and youth addiction to uncontrolled AI agents—but neither wants to slow development and surrender a strategic advantage.

The discussion examines how U.S. semiconductor export controls have constrained China’s access to computing power while also accelerating the growth of its domestic technology industry. It also challenges the American argument that regulating AI would necessarily weaken innovation: Chinese AI companies operate under extensive regulation and continue to advance.

Rather than expecting a sweeping bilateral agreement, New America senior fellow Sam Sachs advocates practical cooperation among technical experts. Informal dialogues and private-sector coordination, such as joint work to patch a major WeChat vulnerability, offer a more realistic path toward reducing risks that cross national borders.

分段落总结

[00:35] Shared AI Risks Amid Strategic Rivalry

[事实] AI safety is a major topic surrounding the meeting between Presidents Trump and Xi Jinping, following incidents and concerns involving rogue AI agents, researcher resignations, and calls to slow frontier development.

[事实] Both countries face vulnerabilities including job displacement, online youth addiction, and the possibility that unmonitored AI agents could disrupt global systems.

[事实] These shared concerns coexist with fears in the United States that a decisive Chinese advantage in AI could fundamentally alter the strategic balance.

[推测] The central obstacle to cooperation is not a lack of awareness about AI risks, but the fear that unilateral restraint would benefit the rival country.

[02:17] The Mixed Effects of U.S. Chip Export Controls

[事实] Beginning under the Biden administration, the United States restricted exports to China of advanced semiconductors and the equipment needed to manufacture them.

[事实] The restrictions curtailed Chinese access to the computing power required both to train AI models and to perform inference.

[事实] In response to reduced access to U.S. technology, China’s domestic semiconductor and AI industries accelerated their development and became less reliant on American suppliers.

[推测] Export controls may therefore be becoming less effective as a long-term technological choke point than U.S. policymakers originally intended.

[03:22] China’s Heavily Regulated AI Industry

[事实] Chinese AI laboratories, companies, and internet platforms have been subject to extensive government regulation for several years.

[事实] In response to concerns about young people becoming addicted to AI companions, China prohibited agents that developed human-like emotional characteristics capable of encouraging addictive behavior.

[事实] Some Chinese AI companies reportedly removed numerous agents because of concerns about regulatory compliance.

[事实] AI regulation in China also serves the Communist Party’s political objectives, including online censorship and preventing models from discussing content blocked by the Great Firewall.

[推测] China’s regulatory model cannot simply be transplanted to the United States because its public-safety measures are intertwined with political control.

[05:12] Regulation and Innovation Are Not Necessarily Opposed

[事实] A common U.S. argument holds that AI regulation would undermine American companies’ ability to innovate and compete with Chinese rivals.

[事实] Sachs notes that Chinese companies remain innovative despite operating under heavy regulation.

[事实] Chinese actors can also point to limited U.S. regulation when questioning why they should restrain their own development.

[推测] Mutual claims about competitive disadvantage risk creating a cycle in which both countries use the other’s behavior to justify avoiding stronger safeguards.

[06:23] Realistic Expectations for U.S.–China Talks

[事实] Sachs does not expect the presidential meeting to produce a grand bargain under which China agrees to slow frontier AI development.

[事实] She argues that the two leaders could instead endorse continued communication among technical experts.

[事实] Informal “track two” dialogues already bring together academics and industry practitioners from both countries to discuss shared concerns and policy.

[事实] Engineering experts can address technical standards and other practical issues that broader political or legal discussions may not resolve.

[推测] High-level political support for expert dialogue may be more achievable—and more immediately useful—than a comprehensive interstate AI agreement.

[08:13] The WeChat Vulnerability as a Model for Cooperation

[事实] A U.S. security company discovered a major vulnerability in WeChat that, if exploited, could have affected hundreds of millions of devices in China and elsewhere.

[事实] The American company contacted Tencent, and the two organizations collaborated to patch and mitigate the vulnerability.

[事实] Sachs presents the incident as an example of private, non-state actors sharing technical information to secure systems whose risks do not stop at national borders.

[事实] Continued cooperation of this kind has become difficult because of current geopolitical tensions.

[推测] Quiet, issue-specific technical collaboration may offer the most durable foundation for AI safety cooperation when formal diplomatic relations are strained.

播客点评/总结

The episode’s chief strength is its rejection of a simplistic choice between AI competition and AI safety. It shows that the United States and China can remain strategic rivals while still sharing a practical interest in preventing technological failures with global consequences.

The discussion also adds useful nuance to perceptions of China’s AI sector. Chinese regulation can address genuine social harms while simultaneously enforcing censorship and political control; meanwhile, continued Chinese innovation complicates claims that regulation inherently prevents technological progress.

[推测] The short format limits deeper examination of whether informal expert exchanges can survive worsening political tensions or influence national policy. Even so, the episode is a useful introduction for listeners interested in AI governance, semiconductor policy, cybersecurity, and U.S.–China technology relations.