Research before regulation

This report presents a map of alternatives to concentrated AI, drawing on thirty-six interviews with the people building and studying them, three case studies written by their builders (OpenMined, the Ethereum Foundation, and Civic AI), and nearly thirty working systems, institutions, and standards. It is written for policymakers, funders, and builders, and sheds light on ways of developing and using AI designed to help people live fuller, freer, and more capable lives. The report presents four findings: 1. Concentration need not be the endpoint of development. 2. The alternatives do not depend on matching the frontier. 3. Chip manufacturing and frontier-scale training remain concentrated. Everywhere else, working systems exist. 4. Most of the alternative sits outside the safety debate that concerns a few frontier systems. Policy recommendations will follow in a separate report in 2027.