Understanding how antibiotics interact with membranes is crucial for predicting their off-target effects, particularly hepatotoxicity. This work compares two clinically important glycopeptide antibiotics, Teicoplanin and Oritavancin, using an integrative approach that combines in vivo pathology, lipid biophysics, drug-lipid interactions by NMR spectroscopy, and molecular dynamics simulations. Despite causing little direct disruption to lipid membranes, Teicoplanin produced significant hepatotoxicity, including increased liver enzymes and histopathological loss. Teicoplanin localises at the membrane-aqueous interface, where it forms stable surface-level interactions that have the potential to periodically disrupt membrane-associated processes. On the other hand, due to its deep insertion into the bilayer core, Oritavancin exhibited a more benign hepatic profile, despite causing stronger membrane perturbation. Long-term cellular stress is probably mitigated by this embedded configuration, which facilitates less interaction with membrane receptors. These findings demonstrate that glycopeptide-induced hepatotoxicity is governed by the topology and duration of membrane interactions rather than simply by their magnitude. The study promotes a lipid-centric framework for the logical development of safer, membrane-active treatments and emphasises the value of lipid membrane models and atomistic simulations as predictive tools in early-stage drug evaluation.