I would like to suggest an AI alignment and governance principle centered on preserving human agency. Most AI systems are commonly described in terms of optimizing for successful task completion, safety, helpfulness, or other reward signals. I believe advanced…