Using intelligent reflecting surfaces (IRSs), wireless propagation channels can be manipulated such that information leakage to eavesdropping terminals in a multiple-input multiple-output (MIMO) setting is significantly suppressed. This observation illustrates the potential secrecy gains of IRS-aided MIMO systems. This work develops a novel low-complexity algorithm by which these potential gains are exploited. Invoking methods from fractional programming, the algorithm iteratively designs the digital precoder at the transmitter and tunes the IRS elements, such that the weighted secrecy sum-rate is maximized. It is shown that as the algorithm iterates, the weighted secrecy sum-rate evolves in a non-decreasing way. Numerical investigations confirm the efficiency of the proposed algorithm.