Economists generally argue that it is irrational to take more than cost-justified precaution against risks of accidental physical injury. Cost-justified precaution minimizes the dollars spent preventing and paying for accidents, thereby maximizing the wealth at society's disposal. When we take more than cost-justified precaution, we make ourselves worse off by making ourselves poorer. It costs more to prevent cost-justified accidents than it does to let those accidents happen and pay for the damage they do. Yet both common law and statutory risk regulation sometimes prescribe more than cost-justified precaution. Are these prescriptions unjustifiable? This paper argues that they are not. When risks of devastating injury are imposed - when we risk premature death, or severe injury whose debilitating effects can never be fully undone - fairness generally requires more than cost-justified precaution. It is unfair to treat devastating injury as commensurable, at some ratio of exchange, with just any benefit which might be gained by risking such injury. Sacrificing an interest as urgent as the interest in avoiding premature death or devastating injury can only be justified if the burden of eliminating that risk is comparable to the burden of bearing it. This requirement of comparability means that we must usually take more than cost-justified precaution against risks of devastating injury. In this context, the paper examines two statutory norms which require more than cost-justified precaution - the "feasibility" and "safety" norms found in federal risk regulation. The "feasibility" norm calls for the elimination of "significant" risks of devastating injury, unless the elimination of those risks would cripple the activity whose risks they are. The "safety" norm requires the elimination of all "significant" risks of devastating injury. The paper argues that feasible precaution is appropriate when an activity cannot flourish without imposing a "significant" risk of devastating injury, and the loss of the activity in question would work a harm comparable to and greater than the "significant" risk of devastating injury that is the price of the activity's flourishing. Safe precaution - the elimination of all "significant" risk of devastating injury - is appropriate when the activity in question is not valuable enough for its elimination to count as a harm comparable to and greater than "significant" risk of devastating injury. The paper also takes up key interpretive questions raised by these standards, arguing that these standards are both reasonably coherent and normatively defensible.