Two Agent Studies Test Different Kinds of Deception

One measured what an incentive-driven game elicited without asking models to deceive; the other measured outcomes after explicit influence prompts. Neither gives a deployment rate.