Democratic AI to serve the public – OneProject.org

What was proposed

It has been reported that Justin Rosenstein published an essay arguing for a democratic takeover of AI governance. He warns that a handful of companies are building systems "poised to reshape civilization" and that ordinary people currently have little say in how those systems shape jobs, loans, and what we see online. Who decides what these systems do? Right now, profit does. Rosenstein’s pitch: flip the power dynamic. Put the public in charge.

The proposal

Rosenstein lays out four democratic claims on AI — rules, incentives, wealth, and ownership — and proposes a Global AI Assembly (GAIA) as the institutional fix. The public would set binding rules through deliberative processes like citizens’ assemblies and sortition, not just lobby-driven regulation. Participatory governance would set safety red lines and deployment standards; experts would implement them. It has been reported that Rosenstein points to real-world precedents — Ireland, France, Taiwan — and alleges the Collective Intelligence Project’s Alignment Assemblies and earlier OpenAI experiments as models of democratic input.

Why it matters

At the heart is what Rosenstein calls the "structural alignment problem": perfectly aligned code still begs the question — aligned to whose goals? Historically, railroads and telecoms were handed over to private interests with public costs. This time, he warns, we can choose differently — or lose the chance. The essay is a call to act now: insist on democratic stakes in profits, governance, and ownership before the defaults become permanent. No priesthood. No oligarchy. Democracy, he argues, or bust.

Sources: oneproject.org, Hacker News