The 1977 National Dam Inspection Program determined that many high‐hazard dams in the United States were incapable of passing a probable maximum flood (PMF). Retrofitting these dams was estimated to cost at least $10 billion. Since the PMF is revised upward periodically, retrofit is a continual issue. But surviving a PMF is a more stringent safety criterion than preventing other sources of dam failure; in addition, it is more stringent than safety criteria for other structures with respect to wind, earthquakes, or storm surges. This higher safety goal has large social costs. We propose an alternative safety goal, separating property damage from possible loss of lives. For a proposed dam whose failure could cause large loss of life or property damage, a careful evaluation should be done as to whether the dam should be built. For dams that impose smaller hazards, property damage should be handled by an analysis based on expected values of annualized benefits and costs. An adjustment for scale could be used if the property damage were extremely large. Danger to lives should be handled by establishing programs to warn and evacuate people. Our proposal should (1) lead to less injury and death, (2) use society's limited resources more efficiently, and (3) put the determination of safety goals on a more scientific and sensible basis.