In a banking model with imperfect information, I find that more precise information increases the economy's vulnerability to bank runs. For low transparency levels, depositors cannot distinguish bad from good states based on their private signals and, absent liquidity shocks, have no incentives to withdraw early. As transparency increases, and private signals become more informative, depositors' incentives to withdraw strengthen and run-proof contracts become costlier in risk-sharing terms: the bank must offer less to early withdrawers to prevent runs. When transparency is high enough, the bank would rather forgo return and hold excess liquidity than choose a run-proof deposit contract.