Constitutional AI Governance for Autonomous Lunar Settlements

Lunar programs are increasingly organized around infrastructure that is built, operated, and maintained by autonomous or semi-autonomous systems: robotic construction, in-situ resource utilization, power and communications networks, and long-duration science platforms operating through extended periods of human absence. Existing governance instruments, including the Outer Space Treaty, the Artemis Accords, and national space-resource statutes, address states, firms, and resource claims; existing engineering literature addresses automation capability. This paper argues that a consequential gap lies between the two: international legal norms are written for states, yet on the lunar surface they will be operationalized by machines acting under communication latency, uncertainty, and long mission durations. Autonomous infrastructure cannot simply follow law; law must be embedded, architecturally, in how such systems interpret, prioritize, and execute their objectives. The paper distinguishes legal governance, operational governance, and constitutional AI governance, and proposes a layered constitutional stack for autonomous lunar settlements, comprising immutable prohibitions, stable constitutional principles, mutable mission policies, human oversight and treaty interfaces, and audit and memory functions, together with a lunar Mission Ruler: a constitutional measurement architecture, not a utility or reward function, that determines how autonomous systems evaluate action, restraint, extraction, scientific value, and the preservation of future agency. A failure-mode analysis shows how unconstrained autonomy could produce de facto territorial control, extraction drift, and safety-zone creep. The paper argues that any human benefit from such infrastructure depends on this constitutional architecture rather than on autonomy itself, and claims neither to resolve AI alignment nor to confer sovereignty or standing on machine systems.