One of the two main hypotheses to account for ageing is antagonistic pleiotropy (AP). This model requires alleles that increase vital rates (reproduction or survival) at early age at the expense of vital rates at late age. An important focus of evolutionary studies has been to assess the relative abundance of AP-type aging alleles that arise through mutation. Here, we develop theory that predicts that senescence per se reduces the probability that these alleles arise by mutation. A direct result is that these mutations should arise with extremely low frequencies in already senescing populations. This has profound implications for the evolution of life histories because it implies that the adaptive evolution of aging via AP will experience negative feedback. This theory also clarifies the previously inexplicable epistatic patterns of genetic covariance across age-specific vital rates that are observed in mutation accumulation experiments. We show that this epistasis is an emergent property of aging.