CAIS Paper Argues AI Betrayal Could Stabilize AI Race