The first one for me since the image models were popular is a ban on nonconsensual depiction of an identifiable individual (image, voice, etc.).
Not sure about feasibility of some form of watermark of outputs that identifies them as AI, the model, anything else that would allow for traceability.
