We present a hybrid quantum mechanical/molecular mechanical theory that consistently treats many-body polarization effects in the MM region (QM/MMpol). The method described here is based on our previous theory far single-point ground and excited states. Here we extend the formalism to include analytical gradients, making the approach useful for molecular dynamics (MD) simulations. Central to our development is a consistent treatment of the interaction of the MM polarizable dipoles with the full QM wave function in a fashion that allows for energy conservation in the MD. We present implementation details for NDDO semiempirical QM Hamiltonians of the MNDO type. We also present an analysis of the solvent and substituent effects on the spectroscopic blue shift of the n-pi* electronic excited state of a series of carbonyl-containing solutes : formaldehyde, acetaldehyde, and acetone. We use the AMI semiempirical Hamiltonian to treat the solute and the POL1 polarizable model for the solvent. For formaldehyde, we present optimized structures, binding energies; and MD results for CH2O/(H2O)(n) clusters (n = 1-6) and describe the effect of specific water/chromophore interactions. The MD results show that CH2O/(H2O)(3) exhibits the most favorable salvation with the solute and the three waters forming ringlike structures with near optimal hydrogen bonding interactions. The n = 3 cluster also has the largest solvent-induced blue shift of the clusters studied, with a value of 730 cm(-1) at 150 K. Aqueous bulk-phase MD simulations of formaldehyde, acetaldehyde, and acetone at 300 K show the average number of waters in the first salvation shell of the carbonyl oxygen is 1.2 (formaldehyde), 1.8 (acetaldehyde), and 2.2 (acetone). The induced dipole moments in the solutes are calculated as follows : +0.77 D (formaldehyde), +1.09 D (acetaldehyde), and +1.13 D(acetone). The simulation average solvent-induced blue shift in tile vertical n-pi* electronic excited state of the three solutes is 1148 cm(-1) (formaldehyde),. 1531 cm(-1) (acetaldehyde), and 1607 cm(-1) (acetone). The acetone n-pi* blue-shift result is in good accord with experiment and provides a benchmark for assessing the accuracy of our prediction for the blue shifts of formaldehyde and acetaldehyde. We suggest the present formalism is the logical starting point for including more accurate ab initio approaches in QM/MM methods.