Rearranging and manipulating highly deformable bags is common in daily life, such as when we put food in grocery bags or papers in backpacks or garbage in trash bags. In contrast to rigid object manipulation, manipulating deformable objects is challenging for robots due to complex object configurations and dynamics. Furthermore, the task of manipulating highly deformable bags introduces extra complexities due to reasoning about 3-dimensional space. In this work, we consider a novel task: grasping and lifting physical bags to contain items. Using a bimanual ABB YuMi robot we test three grasping methods, where positions are determined via a human, a random baseline, and a baseline based on grasping the leftmost and rightmost points of the bag (``Maximum Width''). Across experiments where the YuMi grasps and lift bags to contain a fabric and a cable, we perform 15 trials for each method. Results demonstrate that the human has the best success rates (14/15), followed by Maximum Width (10/15) and then random (5/15).