Authority, Refusal, and Resilience in Autonomous Systems
Year: 2026
Type: book
Abstract: Authority, Refusal, and Resilience in Autonomous Systems develops the Stable Authority Boundary (SAB) as a pre-execution design invariant for autonomous systems, distributed systems, AI-enabled systems, agentic systems, cyber-physical systems, resilient infrastructure, and safety-critical computing. It addresses a fundamental systems-engineering distinction between technical capability and legitimate execution authority: a machine may remain fully capable of performing an action after the conditions that authorized that action have changed, degraded, expired, or disappeared. The work argues that consequential execution is legitimate only when the applicable authority can be identified, authenticated, current, valid, within scope, and sufficiently verifiable before action begins. Where authority is absent, expired, revoked, ambiguous, degraded, out of scope, stale, conflicting, or otherwise unverifiable, the engineered response should be authority contraction, refusal, hold, silence, degraded operation, or another explicitly governed non-action state rather than unrestricted autonomous continuation. This framework treats authority, refusal, silence, resilience, and authority contraction as execution-level engineering properties rather than solely policy, cybersecurity, or post-hoc governance concerns. It examines autonomous decision-making, machine authority, execution authority, authorization boundaries, authority loss, authority drift, degraded coordination, distributed consensus, fail-safe behavior, fail-secure behavior, fault tolerance, graceful degradation, resilient systems, and autonomous-system safety. The central question is not simply whether a system can continue operating, but whether it still may perform a particular action under the authority presently available to it. The distinction is especially important in distributed autonomous systems, edge computing, AI agents, multi-agent systems, robotics, unmanned systems, critical infrastructure, mission systems, and machine-to-machine coordination, where technical availability may survive the loss of communications, supervisory control, network connectivity, upstream authorization, trusted state, or other conditions upon which legitimate execution depends. Conventional high availability and fault tolerance can preserve capability while unintentionally preserving actions whose original authority is no longer valid. SAB addresses this gap by making authority itself an explicit system state and execution constraint. The work also examines resilience as preservation of legitimate purpose rather than uptime alone. A resilient autonomous system must be able to distinguish between capability preservation and authority preservation, contract its permitted behavior when evidence becomes insufficient, refuse actions outside the remaining authority envelope, and recover or expand authority only through governed and verifiable transitions. This connects system resilience, autonomous recovery, degraded-mode operation, failover, reconstitution, state reconciliation, authority restoration, trust restoration, and safe reintegration to an explicit authority model. A related concern is the evidence upon which autonomous execution depends. Modern systems increasingly consume information from sensors, services, software components, external repositories, databases, APIs, AI pipelines, automated workflows, and other machines. Consequently, machine-readable provenance, data provenance, dataset provenance, provenance metadata, data lineage, data integrity, metadata integrity, metadata generation, data validation, source validation, persistent identifiers, traceability, chain of custody, and computational reproducibility can become relevant to execution authority. Information availability does not establish that information remains sufficiently trustworthy, current, or authorized for a proposed action. The same problem applies to automated research and data-processing environments. A system may receive an open research dataset, sample dataset, source dataset, derived dataset, transformed dataset, research data archive, software artifact, simulation result, validation artifact, or machine-generated record. It may encounter a data-processing example, automated transformation, computational workflow, data pipeline, dataset DOI, data DOI, persistent identifier, or versioned research object. The continued existence or technical accessibility of that artifact does not by itself establish its provenance, integrity, applicability, freshness, or authority for subsequent machine action. SAB therefore treats provenance, validation, integrity, identity, scope, currency, and authority as related evidence questions. Tamper-evident data, machine-readable records, authenticated metadata, reproducible processing, persistent identifiers, DOI relationships, version histories, transformation histories, and traceable derived artifacts can help establish what information was used, where it originated, how it changed, which system or actor produced it, and whether an autonomous decision remains supportable. The framework also addresses the problem of semantic drift and authority drift. In long-lived autonomous and distributed systems, the meaning, scope, or operational context of an instruction, policy, resource, dataset, capability, or authority grant can change even when the underlying digital object remains technically available. Explicit authority boundaries, provenance, versioning, validation, and governed state transitions provide mechanisms for detecting when continued execution can no longer be justified from the available evidence. These issues connect the Stable Authority Boundary to broader research in systems engineering, autonomous systems engineering, AI safety, AI governance, distributed computing, resilient computing, dependable systems, fault-tolerant systems, zero-trust architecture, cyber resilience, mission assurance, safety engineering, trustworthy autonomy, machine governance, data governance, research data management, open science, reproducible research, repository analysis, and technical literature analysis. SAB provides an execution-oriented architectural mechanism for determining when a capable machine should act, when its authority should contract, and when it should refuse. This revised edition aligns the monograph with the Stable Authority Boundary (SAB) v1.0 — Conformance Specification, CSI SAB-STD-v1.0-DF-2026, including the practical SAB review framework, eight conformance dimensions, evidence-bearing operation, and hostile falsification audit. Together, the monograph and specification describe an architectural approach in which capability answers what a machine can do; authority determines what it may do; evidence determines whether that authority can presently be relied upon. Stable Authority Boundary (SAB) v1.0 — Conformance Specificationhttps://doi.org/10.5281/zenodo.21939806 BLOCK VECTOR Research Maphttps://doi.org/10.5281/zenodo.22102620 Interactive BLOCK VECTOR Research Maphttps://blockvectortech.com/research-map.html Author ORCID — David B. Forbeshttps://orcid.org/0009-0003-8833-3853
Author: Forbes, David
Language: English
Topic: Access Control and Trust
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7204159875