MAED: Governing the Right to Act in Decision-Capable AI Systems

As artificial intelligence systems increasingly operate in domains where their outputs implicitly authorize real-world action—such as financial allocation, medical decision support, autonomous control, and human–machine interaction—the dominant safety paradigms reveal a critical gap. Existing approaches emphasize correctness, uncertainty signaling, or alignment with user intent, yet fail to address a more fundamental question: when is an AI system epistemically permitted to act at all? This paper introduces the Minimum Acceptable Evidence Doctrine (MAED), a governance framework that regulates the right to action rather than the quality or fluency of responses. MAED enforces a strict separation between analytical capability and decision legitimacy by requiring a minimum threshold of epistemic sufficiency before any action-enabling output is allowed. When this threshold is not met, the system must refuse action, terminate reasoning, or downgrade its output to a strictly non-actionable form. MAED is presented as an operational extension of the Cognitive Reflexivity Function (CRF), which defines conditions under which continued reasoning becomes epistemically illegitimate. While CRF governs when cognition must halt, MAED governs what minimal, non-deceptive information—if any—may be disclosed without conferring false authority to act. Together, they form a layered governance architecture for decision-capable AI systems. We further describe three conceptual enforcement primitives—Honesty Boot, Right to Refuse, and Hard Termination—as first-class cognitive outcomes rather than failure modes. Using high-stakes decision environments as illustrative contexts, we argue that refusal, silence, and termination are not signs of reduced intelligence, but necessary features of responsible artificial cognition. This work reframes AI safety away from optimizing answers and toward governing action under uncertainty, offering a transferable doctrinal foundation for AI systems operating where premature or illegitimate action carries irreversible consequences.