Open Source, Open Science & AI

AI development is consolidating fast: frontier compute concentrated in a handful of labs, research agendas shaped by a few institutions, talent and infrastructure clustering in narrow geographies. This talk examines these centralization pressures and makes the case that open source and open science are not just nice-to-haves but load-bearing for both AI progress and AI safety: open models enable independent safety research, reproducibility, and red-teaming that closed systems can't deliver alone, while open scientific ecosystems keep the field's epistemics healthy and contestable. Drawing on Foresight's AI Nodes -a decentralized network of local hubs for AI research and coordination- let's explore concrete strategies for defending open ecosystems from closure pressures, whether driven by commercial incentives, regulatory capture, or AI systems themselves accelerating winner-take-all dynamics.