The grip force dynamics during grasping and lifting of diversely weighted objects are highly informative about an individual's level of sensorimotor control and potential neurological condition. Therefore, grip force profiles might be used for assessment and bio-feedback training during neurorehabilitation therapy. Modern neurorehabilitation methods, such as exoskeleton-assisted grasping and virtual-reality-based hand function training, strongly differ from classical grasp-and-lift experiments which might influence the sensorimotor control of grasping and thus the characteristics of grip force profiles. In this feasibility study with six healthy participants, we investigated the changes in grip force profiles during exoskeleton-assisted grasping and grasping of virtual objects. Our results show that a light-weight and highly compliant hand exoskeleton is able to assist users during grasping while not removing the core characteristics of their grip force dynamics. Furthermore, we show that when participants grasp objects with virtual weights, they adapt quickly to unknown virtual weights and choose efficient grip forces. Moreover, predictive overshoot forces are produced that match inertial forces which would originate from a physical object of the same weight. In summary, these results suggest that users of advanced neurorehabilitation methods employ and adapt their prior internal forward models for sensorimotor control of grasping. Incorporating such insights about the grip force dynamics of human grasping in the design of neurorehabilitation methods, such as hand exoskeletons, might improve their usability and rehabilitative function.