In human-machine interfaces with joint/link compliance, human hand dynamics can have a detrimental effect on system performance. This paper investigates the effect of human dynamics on the effectiveness of a virtual fixture implemented on an admittance-controlled robot with joint compliance. Two open-loop control methods are proposed for creating a virtual fixture that prevents the user from entering a forbidden region: one that compensates for hand dynamics, and one that predicts overshoot based on the user's current velocity. The methods determine a "safe" location of the virtual fixture that prevents the user from entering the true forbidden region. A first experiment was conducted to determine hand dynamic parameters (mass, damping, and stiffness) and evaluate their relationship to the force applied by the user on the interface. A second experiment demonstrated that both methods for creating safe virtual fixtures prevent the user from entering the forbidden region, even when the interface has significant joint compliance.