The current financial crisis testifies that the sophisticated risk management models used by large financial institutions are inadequate. The main objective of this research project is to analyze the sources of this failure and to develop sound conceptual principles for founding new risk management methods for financial institutions. In spite of the wide use of sophisticated risk management models by the majority of large firms, the conceptual foundations for them are weak. Most of them rely on the assumption that financial markets always function well. The few theoretical models that incorporate endogenous financial frictions use contract theoretic tools but they are static or two period models. Such models cannot generate really testable implications, or provide quantitatively reasonable policy recommendations. Another strand of the theoretical literature has developed diffusion models for modelling the financial behaviour of corporations in continuous time. However this literature is mathematically oriented and makes very strong assumptions, without clear justifications. Our objective is to combine these two approaches and construct testable dynamic models with endogenous financial frictions. These models are to be simple enough that they can provide reasonable policy recommendations, with a particular attention to banks and insurance companies. By adapting the general model of corporate risk management in a dynamic set-up to the specificities of financial intermediaries, we will develop a model of risk management for the financial sector. Implications will be derived for prudential regulation of financial intermediaries and the organisation of supervision, with a particular attention to the prevention and management of future financial crises.