Autonomous systems increasingly operate continuously, adapt dynamically and interact with evolving environments throughout extended operational lifecycles. This paper introduces the concept of Governability by Design and argues that the long-term challenge of autonomous systems may not be achieving compliance at a specific point in time, but preserving the conditions that make compliance continuously achievable throughout runtime operation. The paper develops a Governability Requirements Framework consisting of Observability, Traceability, Controllability, Auditability, Accountability Support, Adaptation Visibility and Predictability. These requirements are positioned as foundational architectural capabilities supporting Runtime Governability and Regulatory Continuity beyond initial conformity assessment and deployment. In addition, the paper introduces the concepts of the Hidden Governability Assumption, Governability as a System Property and the Governability Gap, describing situations in which autonomous systems may remain formally compliant while simultaneously exhibiting insufficient governability during runtime operation. The paper explores how these concepts relate to the objectives of the EU AI Act and argues that future autonomous systems may increasingly require architectural characteristics specifically designed to preserve governability over time. Keywords: Governability, Governability by Design, Continuous Compliance, Regulatory Continuity, Runtime Governability, EU AI Act, Autonomous Systems, Governance Infrastructure, Accountability, Transparency, Human Oversight.
Andreas Blumer (Mon,) studied this question.