Potential US-China AI Treaty to Curb Misuse Is Under Discussion
Policymakers are debating the possibility of a US-China treaty focused on artificial intelligence. The goal of such an agreement would be to curb the proliferation of AI misuse. The discussions reflect growing international concern over AI safety and the need for cross-border cooperation on regulatory frameworks.
- The first intergovernmental U.S.-China dialogue on AI was held in Geneva in May 2024, following up on a commitment made by Presidents Joe Biden and Xi Jinping in November 2023. While the talks didn't produce a formal agreement, both sides discussed AI risks and governance, with the U.S. noting that its technology protection policies were not negotiable. - A key point of contention is U.S. export controls on advanced semiconductors, which Beijing views as a measure to suppress its AI industry. U.S. officials, meanwhile, have raised concerns about China's "misuse" of AI and the potential for deepfakes to influence elections. - China has been actively shaping its domestic AI regulatory landscape, not through a single comprehensive law, but via targeted regulations for specific applications like generative AI and deep synthesis, overseen primarily by the Cyberspace Administration of China (CAC). This approach aims to manage risks while maintaining flexibility to foster innovation. - In October 2023, China introduced its Global AI Governance Initiative, advocating for a UN-centered approach to governance and emphasizing the development rights of all countries, particularly those in the Global South. This initiative is seen as an effort to counter U.S. influence over global AI safety standards. - The Chinese AI agent market is experiencing rapid growth, with the user base for generative AI reaching 250 million by February 2025. Major players like Baidu, Alibaba, and Tencent are competing for dominance, with general AI assistants like Doubao and DeepSeek emerging as major portals. The overall China AI agents market is projected to grow at a CAGR of 50.8% between 2026 and 2033. - Both nations were signatories of the 2023 Bletchley Declaration, which acknowledged the potential for serious harm from advanced AI and the need for international cooperation. However, at a subsequent military AI summit in Spain, neither the U.S. nor China signed a declaration on responsible AI use in warfare, highlighting ongoing strategic competition. - Recent research from Chinese institutions highlights a focus on multi-agent systems (MAS). Papers from the Chinese Academy of Sciences and Peking University explore topics like understanding agent importance via counterfactual reasoning and the mathematics of multi-agent learning at the intersection of game theory and AI. Another paper from Tencent AI Lab and others introduces SciAgent, a unified multi-agent system for scientific reasoning. - While U.S. frontier models have been predominantly closed-source, leading Chinese AI labs like DeepSeek and major firms such as Alibaba and Baidu are increasingly embracing an open-source strategy for their models. This approach is seen as a way to accelerate adoption and gain traction, particularly in the Global South.
