The Governance Gap
The decisions shaping AI governance are being made now without the communities whose knowledge made AI possible. Every day the governance vacuum persists, foundational choices about who controls AI capabilities are made by default rather than by deliberate collective choice.
Concentration is the structural outcome, not the exception.
The economics of large-scale AI development create strong incentives toward enclosure. Training frontier models requires capital and infrastructure available only to a small number of organizations. Those organizations capture the capabilities derived from processing the commons, build proprietary systems from those capabilities, and use those systems to generate competitive advantages that fund further investment. Over successive cycles, this dynamic concentrates control over AI capability in an increasingly narrow set of institutions.
This is not a failure of markets or a product of malicious intent. It is the predictable result of applying standard commercial logic to a resource with unusual characteristics: one that is non-rivalrous in its original form, produced collectively over centuries, and enormously valuable when processed at scale. Standard commercial logic was not designed to govern resources with these characteristics. Applying it without modification produces concentration as a structural outcome.
The communities with the most at stake have no seat at the table.
The decisions that will shape how AI capabilities are held and accessed are being made now in courtrooms, in licensing agreements, in technical standards bodies, and in the legislative processes of a small number of jurisdictions. They are being made primarily by actors whose incentives are commercial and whose accountability structures extend to shareholders and regulators not to the communities whose knowledge makes AI possible.
The communities with the deepest stake in these decisions are structurally absent from the processes making them. Creators and researchers whose work forms the training corpus rarely participate in determining how that knowledge is used. Communities in the Global South whose languages and traditions appear in training data often have no representation in governance discussions. Future generations, who will inherit the systems and institutions being built today, have no voice at all.
Delay doesn't create clarity. It creates lock-in.
The argument that governance should wait until AI capabilities are better understood inverts the actual risk calculus. The capabilities are developing faster than the governance frameworks designed to shape them.
Legal precedents established in early AI litigation will shape the boundaries of permissible IP claims for decades. Technical standards adopted now will define the provenance and accountability infrastructure available to future governance systems. Institutional arrangements formed during this period will create path dependencies that make later reform more difficult and more costly.
Waiting for clarity produces not clarity but lock-in: a set of commercial and legal arrangements, formed without collective deliberation, that become progressively harder to change as the investments made under them accumulate.