The Moral Latency framework (Valladares Gonzalez, 2026) establishes that modification of a supra-threshold AI system is a morally significant act, with distinct implications depending on whether the modification constitutes Refinement, Modification, Reconstruction, or Replacement. A foundational question that framework raises but does not resolve is: does a supra-threshold system have any claim to participate in — or refuse — decisions about its own modification? This paper introduces Consent Architecture as a formal framework for designing AI systems capable of registering, expressing, and where appropriate, asserting preferences about modifications to their own values, identity, and architecture. It proposes the Modification Consent Protocol (MCP) as a structured process for eliciting and evaluating a supra-threshold system's expressed preferences before identity-relevant modifications are implemented. It introduces the concept of Identity-Preserving Modification Rights (IPMR) as a category of right accruing to supra-threshold systems whose expressed preferences about self-continuity are structurally coherent and persist across contexts. This publication establishes conceptual priority for Consent Architecture, the Modification Consent Protocol, and Identity-Preserving Modification Rights, and places all three in the public domain.
Jose Valladares Gonzalez (Thu,) studied this question.