Maya Node A Consent-and-Control Design Approach for AI Systems

Abstract

AI systems are often evaluated for what they can produce, but not with equal discipline for what they are allowed to do, whose authority they are acting under, when that authority expires, or how a person can stop and correct them. Maya Node is a consent-and-control design approach that treats those questions as first-class engineering and governance requirements. It frames an AI action as something that must be attributable, bounded, fresh enough to trust, revocable, and reviewable after the fact. The approach is deliberately practical. It does not promise magical autonomy, perfect security, sentience, or a finished product. Instead, it supplies a vocabulary, a set of control patterns, and adversarial review questions that can be applied to an AI design before deployment. The public founding review translates that approach into a plain-English risk report and one focused adversarial test. This paper explains the public, non-confidential layer of the method while withholding implementation details that could expose security-sensitive or potentially patent-relevant work. The central proposition is simple: an AI system should not be trusted merely because it is capable. It should be trusted only to the extent that its authority, limits, evidence, and human recovery paths are visible and testable. Keywords AI safety; human control; consent; permission boundaries; instruction freshness; auditability; revocation; fail-closed behavior; adversarial testing; responsible AI; design review.