Who decides what problem AI Governance must Solve
AI governance is often discussed as though everyone agrees on the problem it is meant to solve. In practice, different groups may be concerned with different things. Governments may focus on growth, national capability and public protection. AI companies may focus on system performance and technical safeguards. Businesses may be more concerned with whether they can challenge a decision, obtain evidence, retain control or continue operating if an AI service changes or disappears.
Who decides what problem AI governance must solve? examines how those different priorities shape the scope of AI governance.
The central argument is that the way the problem is defined determines what governance is expected to protect. If the main concern is unreliable output, the response will concentrate on accuracy. If the concern is unwanted action, attention turns to permissions and intervention. If the concern is dependence on a powerful provider, the business also needs usable records, practical alternatives and a way to keep operating when access changes.
The book looks at the role of governments, AI developers, standards bodies, independent assessors and businesses themselves. Each has useful knowledge and legitimate interests, but none sees the problem from exactly the same position.
It also examines the limits of narrow assessments. A system may perform its intended task accurately while still leaving important questions unresolved. A business may be unable to challenge an outcome, reconstruct what happened after an update, move work to another provider or maintain essential operations if access is withdrawn.
The discussion draws on the AI Protection Tests, the Standardized Definition of AI Governance, the Evidential Resilience Ratio and the External AI Withdrawal and Geopolitical Disruption Test to examine control, accountability, evidence and dependency from different perspectives.
The underlying question is simple: before deciding whether an AI governance framework is adequate, businesses need to understand what problem that framework was designed to solve and what problems it leaves outside its scope.