Compute & Control: The Physical & Policy Foundations of AI Governance
A joint talk by Naci Cankaya (MIRI) and Haakon Huynh (Columbia University).
Wednesday, May 13th 2026 · 18:00–19:30 CEST (Online)
Event Summary
How do you govern a technology when its physical infrastructure spans nations that might view oversights as threats, and when the people building it may be the only ones who can verify what they're doing?
This event brings two researchers working on two ends of the same problem into one conversation:
Naci Cankaya (Technical Governance Team at MIRI) works on the verification mechanisms, hardware monitoring, and confidence-building tools that could make claims about AI being testable, domestically and across borders.
Haakon Huynh (Columbia SIPA) works on the geopolitical layer. This includes multinational coalitions and infrastructure policies focused on global compute governance.
Together, they'll examine what enforceable AI governance actually requires.
Schedule
18:00–18:05 Introduction and framing
18:05–18:25 Naci on verification mechanisms for low-trust AI governance.
18:25–18:30 Top-voted questions
18:30-18:50 Haakon on compute infrastructure and multinational governance.
Rest of time: Audience Q&A (vote on questions via Slido link)
🌟 For more details on Naci and Haakon's presentations, see a short blurb for each here, in the speakers' own words :)
About the speakers
Naci is a technical AI governance researcher at the Machine Intelligence Research Institute (MIRI), and an affiliate of the Oxford Martin Hardware AI Governance Lab (HAIGL). He worked on transparency and verification mechanisms at MATS (particularly on technical solutions for monitoring untrusted AI datacenters), in collaboration with researchers at RAND and the Future of Life Institute. At MIRI, he conducts and coordinates research on the assurance and confidence-building mechanisms for multinational agreements on safe and restrained AI development.
Haakon is an AI policy researcher focusing on global compute governance, working with the Oxford Martin School AI Governance Initiative and the Future Economies Kuwait Research lab at Columbia School of International and Public Affairs (SIPA). He draws on experience in international diplomacy at the Norwegian Ministry of Foreign Affairs and the United Nations in Geneva, grounding his policy recommendations in practical feasibility.
Who should attend?
AI governance researchers, policy professionals, technical AI safety researchers, and anyone questioning how international coordination on AI could actually work in practice. Open to all. No prior expertise required.
Hosted by
Safe AI Germany (SAIGE) — building Germany's infrastructure for safer AI.