Optimized for mortality: AI, accountability, and the moral future of the intensivist

Artificial intelligence is rapidly entering critical care medicine, not through formal implementation programs, but through the daily habits of trainees and clinicians who are already using these tools informally. The standard framing of this transition, augmentation rather than automation, physician as co-pilot, is reassuring but incomplete. It does not ask the harder question: what does physician responsibility mean when the cognitive work of medicine is increasingly performed by systems the physician did not build, cannot fully interrogate, and may not be trained to override? We argue that the coming decades will not simply require clinicians to learn new skills but will require the profession to confront a genuine moral problem: whether it is possible to bear authentic accountability for decisions you did not make. We pose this not as a question with a ready answer, but as one the profession must begin to sit with before the trajectory of AI adoption settles it by default.