US-China AI rivalry needn’t preclude cooperation

Artificial intelligence (AI) carries risks that transcend national boundaries, encompassing cyber threats, misinformation, system failures, and potentially hazardous misuse. In an increasingly polarized global landscape, does greater divergence among nations enhance AI safety? The United Nations Secretary-General, Antonio Guterres, delivered a resolute response at the inaugural United Nations Global Dialogue on AI Governance: "When countries establish common tests, risk measurements, and accountability frameworks, safety travels alongside the technology."

Conversely, "a disjointed array of incompatible rules elevates costs, isolates the global community, and yields no protection." Yet, as AI necessitates unprecedented collaboration, geopolitical dynamics jeopardize global unity. China launched the World AI Cooperation Organization (Waico) last month, inviting 29 founding members, underscoring its commitment to cooperation and AI accessibility for developing nations.

Meanwhile, the United States has crafted the Pax Silica initiative, focusing on AI, technology, and supply-chain security. While both initiatives intersect, they do not target the same challenges. Waico aims to foster development and global cooperation, fortifying China's global AI governance influence; Pax Silica seeks to tackle genuine concerns about critical technologies and supply-chain resilience while constructing a US-led network of trusted allies.

Given that these initiatives stem from formidable rivals, attributing their essence solely to geopolitical rivalry is misleading. For instance, Indonesia concurrently joined Waico and Pax Silica discussions; Kazakhstan participated in both. Could international AI cooperation perpetuate geopolitical factions? Geopolitical tensions arise from commercial, strategic, and political interests, but perceptions play a crucial role.

Once a nation is perceived primarily as a strategic adversary, cooperation may be misconstrued as bloc formation, regulation as political control, and ambiguous actions as hostile intent. This matter is not purely theoretical. A report by the Carnegie Endowment for International Peace cautions against dismissing Chinese AI regulations due to the Communist Party's control or viewing them exclusively through geopolitical competition.

The reality is nuanced: political and information control play a role, but policymakers are also responding to AI's societal, economic, and ethical implications. Geopolitical differences can coexist with policy convergences. A 2025 study of US and Chinese AI governance, analyzing over 40 policy and corporate governance documents, revealed overlaps in system reliability, algorithmic transparency, and employing AI for safety enhancement.

The analytical error lies in conflating political motives with an exhaustive explanation. Consider human oversight. At the World AI Conference in Shanghai, Chinese President Xi Jinping advocated for human control over AI while emphasizing risk prevention and opposing abuse. Just two weeks prior, the UN Global Dialogue on AI Governance listed "robust human oversight" as a governance priority.

Although the language may seem similar, it does not necessarily imply identical motivations, objectives, or implementation mechanisms between the Chinese and UN approaches. However, institutional differences do not necessarily imply fundamentally distinct policy concerns. To comprehend a policy, we must examine what problem it aims to solve and how it operates within its institutional framework.

A country's political system should be understood within its historical and cultural context, rather than being judged against preconceived notions about what the system should be, thereby pre-determining its meaning. China's Communist Party leadership is an integral part of its institutional environment. Acknowledging its significance is different from using it as an all-encompassing explanation.

An AI policy may reflect political considerations, technological safety, economic development, national security – or multiple objectives simultaneously. Journalists, think tanks, and academics must refrain from minimizing genuine differences or fabricating common ground. Instead, they should accurately identify both and distinguish evidence from inference.

The decisions of Indonesia and Kazakhstan suggest a viable alternative: participation without alignment. Countries can collaborate within different technological ecosystems while preserving their independence in policy-making. Forcing them to choose may exacerbate the very divisions global AI governance aims to overcome. The Chinese phrase "he er bu tong" encapsulates harmony without uniformity.

Countries shaped by distinct histories, cultures, and circumstances do not necessarily need to share political systems, values, or strategic interests to cooperate. As AI will not wait for geopolitical rivalries to diminish, governments must collaborate across differences to ensure AI serves humanity rather than exacerbating geopolitical divides.

Those who analyze international affairs have a responsibility: to acknowledge disagreements without magnifying them through assumptions and to recognize common ground where it exists. These perspectives reflect the author's own views.