Longevity risk is an important driver of wealth accumulation, which influences the dynamics of inequality within and across generations. How should redistributive policies optimally share longevity risk? This paper sheds light on this question by developing and quantifying a dynastic Mirrlees model with idiosyncratic uncertainty in labor productivity and longevity. My analysis concentrates on a largely overlooked fiscal tool for providing insurance against lifespan risk: taxes on wealth, in the form of savings and bequests. I derive novel formulas linking optimal taxes on savings and bequests to the degree of annuity market imperfections at the optimum. In the presence of a positive correlation between productivity and life expectancy, optimal annuity markups are positive in expectation, creating a force for savings taxation and for bequest subsidization. The latter revisits the conventional view whereby imperfect annuitization rationalizes high taxes on bequests. When calibrating the model using US data, I find that the forces of differential longevity on optimal wealth taxes are substantial, and may lead to significant welfare gains.