China's AI Safety Evolution: From National Priority to Global Imperative
As Artificial Intelligence capabilities rapidly advance and Chinese models close the gap with the leading edge, understanding China's approach to AI safety and governance has become more critical than ever. This podcast delves into the "State of AI Safety in China (2025)" report, offering a comprehensive look at how the nation addresses general-purpose AI risks, especially dangerous misuse, accidents, and the profound challenge of losing control over advanced AI systems.Join us as we explore five key domains of China's evolving AI safety landscape from May 2024 to June 2025:• Domestic Governance: Discover how AI safety has been formally elevated to a national priority, with increasingly prominent calls for risk mitigation from top officials, including at the Third Plenum and a Politburo study session. We'll discuss China's expanding AI standards system that elaborates on existing regulations, even revealing intentions to address severe risks like AI's impact on cybersecurity or loss of control.• International Governance: Learn about China's emphasis on AI safety and global AI capacity-building as key themes in its international diplomacy. Hear how Chinese officials have cautioned against unchecked international competition, referring to it as a "gray rhino" risk. We'll also cover the launch of the "China AI Safety & Development Association" (CnAISDA) as China's counterpart to other national AI safety institutes, positioning itself to facilitate international dialogue on extreme risks.• Technical Safety Research: Uncover the rapid expansion of frontier AI safety research output in China, more than doubling in the last year. Previously underexplored areas such as alignment of superhuman systems and mechanistic interpretability have become popular topics, alongside research into deception, unlearning, and CBRN (chemical, biological, radiological, and nuclear) misuse.• Expert Views on AI Safety and Governance: Explore how expert discourse in China is placing greater emphasis on AI safety and governance, reflected in increased coverage at major AI conferences like the World AI Conference (WAIC), which was upgraded to a "High-Level Meeting on Global AI Governance". Experts are also publishing in-depth analyses of AI risks in biosecurity, cybersecurity, and open source AI, often framing AI as a "double-edged sword".• Industry Governance: Understand the impact of leading Chinese foundation model developers signing voluntary "AI Safety Commitments," pledging measures like dedicated safety teams, red teaming, and investment in frontier safety research. We'll also examine the current state of transparency, noting that while companies implement well-known safety methods, detailed safety evaluation results for severe risks are often limited in public disclosures.This podcast highlights China's critical participation in the global effort to govern advanced AI systems, emphasizing that mutual understanding is the foundation for coordination, especially in geopolitically complex times.
Italia