Two current issues in management of public risks, ambiguity and learning, are addressed in the context of managing ecosystems with thresholds, and regulating treatment safety as might apply, for example, to human and animal health, pesticides and herbicides. Reconsidering a recent claim that, in systems that penalise violation of thresholds, learning opportunities induce riskier decision-making, I find no incentive for ambiguity seeking. But opportunity to benefit from learning may indeed induce riskier decisions, an effect that diminishes and eventually disappears as penalties become larger. A recent claim that a rational regulator of pharmaceutical drugs would be ambiguity preferring – a claim that has obvious applicability to a broader set of treatments – is then examined. Ambiguity-tolerant patients may indeed prefer a menu of ambiguous treatments and opportunity to learn and switch, rather than a single treatment with known risk. But the source of ambiguity matters. Patient heterogeneity, prior to and independent of policy, generates ambiguity for individuals and motivates preference for a menu of treatments. However, expanding the menu does not justify approving treatments that are generally riskier. Finally, I challenge the perennial claim that the regulator of risks to human health and safety should seek to maximise expected value.