The Diverging Paths of AI Governance: A Look at China and the U.S.

In a striking juxtaposition, just three days after the Trump administration unveiled its eagerly awaited artificial intelligence (AI) action plan, China responded with its own AI policy blueprint. This wasn’t mere coincidence. On July 26, coinciding with the commencement of the World Artificial Intelligence Conference (WAIC) in Shanghai, China rolled out its “Global AI Governance Action Plan.” This event, the largest AI gathering in China, attracted notable figures such as Geoffrey Hinton and Eric Schmidt from the Western tech realm.

The atmosphere at WAIC stood in stark contrast to the Trump administration’s America-first, deregulated vision of AI. During the opening ceremony, Chinese Premier Li Qiang emphasized the critical need for global cooperation in the burgeoning field of AI. Following him, leading Chinese AI researchers presented compelling discussions that highlighted pressing questions largely overlooked by U.S. policymakers.

Safety Takes Center Stage in China’s AI Dialogue

Among the highlights was Zhou Bowen, head of the Shanghai AI Lab—one of China’s foremost AI research institutions. Zhou addressed the importance of AI safety, suggesting that the government could play a significant role in monitoring commercial AI models for vulnerabilities. This message gained support from Yi Zeng, a prominent AI scholar at the Chinese Academy of Sciences, who advocated for international collaboration among AI safety organizations. “It would be best if the UK, US, China, Singapore, and other institutes come together,” Zeng remarked in a WIRED interview.

The conference also hosted private discussions focusing on AI safety policy. Insights from Paul Triolo, a partner at the advisory firm DGA-Albright Stonebridge Group, indicated that these discussions were fruitful, albeit without any American representation. “With the US out of the picture, a coalition of major AI safety players, co-led by China, Singapore, the UK, and the EU, will now drive efforts to construct guardrails around frontier AI model development,” Triolo noted.

Interestingly, many Western attendees were taken aback by how much the dialogue in China centered around safety regulations. “You could literally attend AI safety events nonstop in the last seven days,” Brian Tse, founder of the Beijing-based AI safety research institution Concordia AI, commented. Earlier that week, Concordia AI had hosted a comprehensive safety forum featuring renowned researchers such as Stuart Russell and Yoshua Bengio.

A Shift in the AI Landscape

When contrasting the Chinese AI blueprint with its U.S. counterpart, it appears that the two nations have flipped their stances. Initially, there was a prevalent belief that Chinese companies developing advanced AI models would face limitations due to stringent censorship. Today, however, U.S. leaders focus on ensuring that local AI models “pursue objective truth,” a phrase criticized by experts for reflecting a top-down ideological bias.

Conversely, China’s AI action plan reads like a manifesto advocating for global cooperation. It recommends the United Nations take a lead in international AI matters and emphasizes the vital role of governments in regulating technological advancements. Strikingly, despite the stark differences in governance styles, both nations share similar concerns about AI safety—model inaccuracies, discrimination, existential risks, cybersecurity vulnerabilities, and more.

Given that both China and the U.S. are developing frontier AI models utilizing similar architectures and scaling methodologies, the societal impacts and risks involved are increasingly congruent. Brian Tse highlights the convergence of academic research on AI safety between the two countries, particularly in areas like scalable oversight and the establishment of interoperable safety testing standards.