Dams Without Foundations

In the AI Story

The critique is not a rejection of dam-building. Toyama's framework insists that governance is essential; without institutional structures, AI is governed by the market alone, which is to say by the interests of those who build and sell the tools. The critique is that governance is itself subject to the Law of Amplification. Institutions with strong governance capacity produce strong governance; institutions with weak governance capacity produce governance on paper that amplifies their existing dysfunction. The EU AI Act is strong because European institutional capacity is strong. National AI strategies in countries with weaker governance capacity are less effective, not because the strategies are worse but because the institutional foundations that would implement them are weaker.

The implication is that governance investment must be paralleled by investment in governance capacity. A regulatory framework without the institutional infrastructure to apply it is a framework on paper. An ethics board without the cultural authority to constrain organizational behavior is a performative exercise. A national AI strategy without the educational and professional infrastructure to execute it is rhetoric. In each case, the governance artifact is real, but the outcomes depend on the institutional capacity to make the artifact binding, and institutional capacity is distributed as unequally as technical capacity.

The global dimension is the most troubling. AI governance is emerging in rich countries where institutional capacity is strong, producing frameworks that protect those countries' citizens. Poor countries with weaker institutional capacity are either exempt from (or dependent on) these frameworks, with the dependencies running through corporate terms of service rather than democratic deliberation. The student in Dhaka who uses Claude Code is governed by Anthropic's policies, not by any framework her country or community has deliberated on. The dam protecting her, if one exists, was built in San Francisco by people who have never been to Dhaka.

Toyama's prescription is structural: governance must be built with the participation of those downstream, not only by those at the source. The technical capacity to build dams is now widely distributed. The institutional capacity and political will to build them democratically, inclusively, and in ways that serve populations without voice at the design stage are not. Building this capacity is the work that the AI era most demands and least provides.

Origin

The framework is Toyama's critical response to Segal's beaver-dam metaphor, developed explicitly in this volume. The underlying analysis draws on Toyama's Law of Amplification applied to institutional governance, and on the broader literature on inclusive institutional design.

Key Ideas

Dams are jurisdictional. Every governance structure operates within a specific scope of authority and leaves the rest of the world subject to the unmediated flow.

Governance is amplified. Strong institutional capacity produces strong governance; weak capacity produces governance on paper.

Corporate governance is bounded by corporate interest. Ethics boards, responsible AI policies, and terms of service protect corporate interests, which may or may not align with broader public interest.

Global asymmetry. AI governance is emerging in rich countries and propagating through corporate channels to the rest of the world, with radical differences in democratic accountability.

Requires participatory design. Governance adequate to the AI era must be built with the participation of those who will live downstream, not only by those at the source.