A method, computer system, and/or computer program product dynamically adjusts an insurance policy parameter for a self-driving vehicle (SDV) operating in manual mode. One or more processors receive a copy of manual control signals from an SDV, where the SDV is in manual mode during a particular time period. The processor(s) also receive a copy of computer control signals generated by an SDV on-board computer on the SDV during the particular time period, and compare the manual control signals to the computer control signals. In response to the manual control signals matching the computer control signals within a predetermined range, the processor(s) adjust an insurance policy parameter for the SDV while the SDV is being controlled by the particular human operator.