Wild animals rarely grow old, and die due to extrinsic hazards. So, natural selection favors genes with good early effects even when they lead to senescence and death at later ages. This implies the Medawar accumulation of late-acting deleterious effects and Williams-Kirkwood life-history trade-off: mortality of a generation strongly depends on its early age life history. Human and protected animal populations live in evolutionary unprecedented conditions, and survive to old age. Quantitative analysis of their mortality establishes that in such conditions a dominant fraction of mortality yields an exact law. In contrast to mortality, the law is biologically non-specific (i.e. independent of genotypes, phenotypes, life history, old age diseases, and all other relevant factors, describing the population and its environment from conception to the age of death). It is universal for species as remote as humans and flies and may lead to the formulation of a biologically non-specific thermodynamic mechanism of mortality. The law predicts that mortality may be significantly decreased and reversed to its value at a much younger age. The reversal is consistent with demographic data. For instance, Swedish females, born in 1916, at 48 years restored the mortality rate they had at 20; Japanese females, born in 1927, at 28 restored the life expectancy they had 8 years earlier. The law quantitatively tests mortality theories and establishes their limitations.