The Trajectory Dean Xue Lan - A Multi-Pronged Approach to Pre-AGI Coordination (AGI Governance, Episode 10)
Oct 3, 2025
Dean Xue Lan, a distinguished chair professor at Tsinghua University, dives into the complexities of AGI governance. He discusses the necessity of an adaptive governance network involving the UN, nations, and companies. Xue emphasizes the emerging roles of national AI safety institutes and the importance of company commitments to safety. He also highlights the need for international coordination, contingency planning for AGI incidents, and building trust between the U.S. and China. Xue's insights pave the way for a nuanced approach to managing AI's challenges.
AI Snips
Chapters
Transcript
Episode notes
AI Is A Regime-Complex Problem
- The 'regime complex' means many overlapping institutions claim governance over AI, creating contradictions and gaps.
- No single hierarchical regime fits AI; governance must span UN bodies, standards organizations, regional blocs, and others.
Governance Needs Overlapping Layers
- Multiple complementary actors are emerging: UN convening, national AI safety institutes, and company safety networks.
- These overlapping layers form parts of a governance ecosystem rather than a single solution.
Create Linked National Safety Institutes
- Build national AI safety research institutes and link them internationally to create shared research agendas and testing standards.
- Use government-supported but networked institutes to convene cross-country technical cooperation on safety.