"Can AI be wrong?" is an editorial investigation into the reliability of conversational AI, centered around a series of documented interactions in which I deliberately pushed ChatGPT to its edges, catching it in contradictions, evasions, and subtle manipulations.