Anti-Money Laundering regulatory frameworks share an unstated design premise: the adversary is human. We demonstrate that this premise is architecturally false: a fully autonomous, deterministic multi-agent system can perform every function of a human criminal organization without possessing computational mens rea. We formalize this system as HYDRA-A and prove that against it, intent-based AML detection mechanisms have zero expected utility. Contributions: (C1) A formal model of the autonomous AML adversary with role-by-role replacement of human criminal organization structure. (C2) The Theorem of Absent Mens Rea: a formal proof that a fully-specified autonomous adversary cannot possess criminal intent, derivable from the typed component interfaces of HYDRA-A. (C3) A Corollary of AML Framework Obsolescence: every intent-based detection mechanism has zero expected utility against HYDRA-A. (C4) A personality vector P=(rho, alpha, sigma, tau) with five adversarial archetypes. (C5) A No Stationary Nash Equilibrium theorem for the arms race between HYDRA-A and adaptive defenses. (C6) Zero-knowledge behavioral verification is the only architectural class that provides a formal defense guarantee against HYDRA-A.
Alejandro Jaime (Sun,) studied this question.