We present an analytical framework for the forward-looking measurement of extreme market risk. In contrast to standard techniques relying on past return data, we propose to extract Value-at-Risk and Expected Shortfall under the physical measure from current option prices. Our empirical evidence suggests that the resulting estimates accurately capture the tail risk of the S&P 500 and that they quickly react to changing market conditions. Compared to dynamic tail risk forecasts driven by past returns, our forward-looking estimates are relatively higher during good times and lower during adverse economic conditions, which could reduce the amplification effects of conventional dynamic risk management policies.