It was observed in THOR 50M vehicle testing that the shoulder belt moved in medial direction toward the neck, slipped over the shoulder pad collar and became entrapped in the gap adjacent to the lower neck load cell. The entrapment caused the unhuman-like response of the THOR 50M, i.e. higher lateral shear load at the lower neck, lower fore-aft load at the left clavicle, and also reduced the chest deflection in upper left and lower right quadrants. The goal of this research is to address the belt entrapment issue without altering the other responses of the dummy. Three design options were analysed with THOR 50M finite element model. The belt at no-slip condition between the original design and the design change options were analysed to compare the output from the dummy sensors. One of the design options demonstrated the belt entrapment issue was addressed when belt slippage occurred. In the meantime, the dummy response of this design remains unchanged when the belt slippage did not occur. The analysis also showed how the other design options would create unfavourable changes to the other responses of the dummy. A prototype part was made for validation testing in vehicle and sled environments, and the results confirmed the observation from the finite element model analysis.