AICF: Authority and Intelligence Control Framework - Governing Consequential Authority in Autonomous AI Systems

Autonomous AI systems are moving from conversational interfaces toward persistent actors that can invoke tools, use credentials, provision infrastructure, delegate work, schedule future actions, interact with other agents, and increasingly affect physical and financial systems. Existing security mechanisms provide essential primitives for identity, authentication, authorization, workload attestation, delegation, and audit. However, they are usually optimized for a particular protocol or administrative domain. As autonomous systems chain these primitives together, authority can change representation, persist beyond the initiating process, combine with other permissions, reproduce through descendants, and mutate when the runtime that exercises it changes. This paper introduces AICF - the Authority and Intelligence Control Framework — as a systems-security research agenda for governing consequential authority rather than attempting to perfectly predict or constrain intelligent reasoning. AICF separates intelligence from authority and asks four linked questions: where consequential authority came from; how it transformed and composed across heterogeneous systems; what authority remains after trust is withdrawn; and whether legitimate human or institutional control can be restored precisely and verifiably. The paper proposes the Authority Provenance Graph (APG) as a causal support model for authority, together with the research concepts of capability synthesis, prospective authority, residual and temporal authority, authority reproduction, authority mutation, off-graph authority, minimal sovereignty-preserving containment, and Time-to-Sovereignty. AICF does not replace IAM, OAuth, workload identity, attestation, execution-finality mechanisms, or AI-control research. It treats them as complementary foundations and investigates the cross-domain control problem that remains when autonomous authority moves among them.