We introduce a conceptual framework for negotiative alignment in high-stakes clinical AI, where human experts iteratively refine AI outputs rather than merely accepting or rejecting them. This approach leverages graded feedback---including partial acceptance of useful insights---to systematically flag and score clinical errors. Although we do not present finalized experimental results, we outline a proof-of-concept using an internal radiograph dataset and a multimodal model (MedRAX) to demonstrate how these severity-scored errors might guide future targeted updates. By grounding each AI-generated report in a continuous, co-adaptive dialogue with clinicians, negotiative alignment has the potential to boost trust, transparency, and reliability in medical diagnostics and beyond.