This article analyzes the last financial crisis focussing on the recurrent dynamics of externalities in banking. It shows that two major determinants of the crisis were the uncertainty of a new form of financial intermediation and the failure of regulation to cope with its externalities. Differently from more standard explanations, which are based on opportunism and/or irrationality of financial institutions, this analysis suggests that regulation has not just been insufficient. Regulation has contributed to financial instability too, supporting fragile conventions to handle uncertainty and encouraging regulatory arbitrage through financial innovation. Three implications are derived from this analysis and contrasted with the ongoing reforms of financial regulation in the US and in the EU. First, regulatory arbitrage is better dealt with by protecting banking from disintermediation than by extending prudential regulation to non-banks. Maturity transformation should be defined functionally and reserved to banks. Second, a major danger with ratings is that they support false conventions of safety. To avoid this outcome, the relevance of ratings for regulatory purposes should be reduced and rating agencies should be deterred from rating without verifiable background knowledge. Third, in order to reduce short-termism, regulation of banks' corporate governance should rather question than reinforce managerial accountability to shareholders.