This paper develops a decision-theoretic framework reconciling the weak forms of the Precautionary Principle (PP) and the Innovation Principle (IP), and applies this to AI innovation governance. It provides formal underpinning for practices in regulatory sandboxing, as well as the limits thereof.