Note: this is an early draft of a proposed governing standard for how AI systems should be constructed and deployed so that humans retain meaningful authority over automated processes. The paper outlines the concept of a "Human Agency Architecture" (HAA), defines its core principles, discusses implementation primitives, and proposes governance and compliance mechanisms. The goal is to provide a practical, engineerable set of requirements that organizations can adopt to ensure AI systems support human judgment, maintain auditability, and enable intervention points where humans can understand, modify, or stop automated actions. Key sections include: definitions and scope; human-in-the-loop and human-on-the-loop distinctions; interface and signaling requirements; logging and provenance for accountability; failure modes and mitigation; compliance checklists; and recommended testing and certification paths. The draft invites feedback from engineers, product managers, policymakers, and civil society to iterate the standard before broader consultation and piloting.