Kyle Chan is a foreign-policy fellow at the Brookings Institution focused on China and AI competition. He is the author of a "multiple races" framework on U.S.-China AI competition, which he set out in a May 14, 2026 interview with Ross Douthat for the New York Times "Interesting Times" series, recorded during the Trump-Xi Beijing summit (China's Not the Problem. We Are. (NYT Interesting Times, Douthat-Chan, May 14 2026)).
"Multiple races" framework
Chan's central argument is that China is not running the same artificial general intelligence (AGI) race as the United States, and is instead running parallel races on efficiency, diffusion, open source, and applications, especially robotics (China's Not the Problem. We Are. (NYT Interesting Times, Douthat-Chan, May 14 2026)).
He points to several indicators in support of this reading. China declined H200 chip access on May 14, 2026 to preserve Huawei's domestic semiconductor program; Chan argues that "if they were really sprinting today for A.G.I., they would've gobbled up those chips as quickly as possible." China's National People's Congress 15th Five-Year Plan, issued in March 2026, designates "embodied intelligence" as a priority, which Chan characterizes as an applications focus rather than an AGI focus. He notes that Chinese tech founders sometimes echo U.S. AGI language, but argues that Beijing itself is "not A.G.I.-pilled" (China's Not the Problem. We Are. (NYT Interesting Times, Douthat-Chan, May 14 2026)).
Risk hierarchy: cyber and bio over AGI
Chan reorders the U.S.-China AI risk hierarchy. He argues that AGI risk "has been overblown," while cyberrisk and biosecurity risk have been "underestimated up until recently" (China's Not the Problem. We Are. (NYT Interesting Times, Douthat-Chan, May 14 2026)).
He positions himself between AI hawks, who favor full-spectrum chip controls, and Silicon Valley accelerationists, who favor unfettered deployment. On Chan's account, existing export controls should be maintained because of cyber and bio risk rather than AGI risk, and the United States should engage China on cyber and bio governance (China's Not the Problem. We Are. (NYT Interesting Times, Douthat-Chan, May 14 2026)).
Position on negotiation
Chan rejects framing U.S.-China AI relations as a binary choice between "race" and "detente." He identifies open-source risks, non-state-actor model arbitrage, and biosecurity as areas worth engaging on, while arguing that binding constraints, verification, and treaties are not yet worth engaging on because trust between the two countries is structurally low. He cautions that "we might be waiting for [an incident] to happen before we take action" (China's Not the Problem. We Are. (NYT Interesting Times, Douthat-Chan, May 14 2026)).
Relationships
- author-of: China's Not the Problem. We Are. (NYT Interesting Times, Douthat-Chan, May 14 2026) — the multiple-races interview
- related: Ross Douthat (interviewer), Brookings Institution AI Initiative (institutional affiliation), US-China AI Competition: Different Races, Different Metrics
- contradicts: AGI-Manhattan-Project framing (Aschenbrenner-style); aspects of 2028: Two Scenarios for Global AI Leadership (Anthropic) (Anthropic's "decisive 12-24-month lead" thesis)