Minimum Sufficient Oversight Principle Formalizes Autonomy Delegation in AI Systems

The Minimum Sufficient Oversight (MSO) principle frames autonomy delegation as a variational problem on the Fisher information manifold, yielding a water-filling allocation of governed delegation across task space. The framework proves a capacity theorem for stationary review policies and derives a first-order approximation linking workflow complexity to quality degradation in multi-model pipelines.