Regulators generally have tried to address the problems posed by the excessive risk-taking of Systemically Important Financial Institutions (SIFIs) by placing restrictions on the activities in which SIFIs engage. However, the complexity of these institutions makes such attempts necessarily imperfect. This article proposes to address the problem at its very source, which is the incentives that SIFI owners have to push for excessive risk-taking by managers. Building on the traditional rule of “double liability,” we propose to modify the current (general) rule limiting the liability of SIFI shareholders to the amount of their initial investments in such companies. We propose replacing the extant limited liability regime with a new system that imposes additional liability over and above what SIFI shareholders already have invested in a pre-set amount that varies with a SIFI’s centrality in the financial network. Our liability regime has a number of advantages. First, by increasing shareholder exposure to downside risk, it discourages excessive risk-taking. At the same time, by placing a clearly defined ceiling on shareholders’ total liability exposure, it will not obliterate shareholders’ incentives to invest in the first place. Second, the liability to which shareholders are exposed is carefully tailored to the level of systemic risk that their institution creates. Thus, our rule induces shareholders to account for the negative externality SIFIs can impose without unduly stifling such financial institutions’ role within the financial system and in the wider economy. Third, as the amount of liability is clearly defined ex ante using the rigorous tools of network theory, our rule minimizes the influence of interest groups and the impact of idiosyncratic government decisions. Last, as markets know in advance the amount of liability to which shareholders are exposed, our rule favors the creation of a vibrant insurance and derivative market so that the risk of SIFIs defaults can be allocated to those who can better bear it.