Hilzoy at the Washington Monthly has a good post regarding VaR's role in the current financial crisis.
NYT starts to get at the problem that VaR created in the current financial crisis
Naked Capitalism points out that VaR is always wrong as it assumes the normal distribution which is not correct for asset prices.
Baseline Scenario points out that the future will not be like the past and both the parameters you use to estimate the impact of an unlikely event as well as the distribution of outcomes will change. You can't use the past to predict the future.
