This Nature commentary examines how responsibility for scientific errors should be assigned as AI systems take on increasingly autonomous roles in research. It cites an AI system built by researcher Chris Lu and colleagues that can generate hypotheses, run computational experiments, interpret results, write papers, and review its own work. The piece argues that accountability for mistakes made by such systems may need to be divided among the users who deploy them, the institutions that host the research, and the developers who build them, rather than resting with a single party.
