As medical AI advances, lawmakers and health leaders face a crucial question: Who is liable when the technology fails?