Authority Engineering Language: A Machine-Readable Contract for Legitimate Action

Paper

21

Authors

A.G. Mauro and C.A. Harris

Date

2026-07-19

Collection

Standing Framework Research

Abstract

Agent systems need more than instructions that say "ask before dangerous actions." They need machine-readable authority contracts: who may decide, which action is at stake, what evidence is required, which scope applies, what obligations constrain the decision, and where escalation goes when authority is missing. This paper describes the Authority Engineering Language, or AEL, as a candidate language for representing those contracts. The local AEL v0.3.5 specification defines syntax, static semantics, runtime action-event evaluation, optional pattern and temporal modules, deterministic audit requirements, and a reference intermediate representation. The adjacent Authority Engineering materials define a compact ontology around Authority Units, Decision Points, Obligation Units, Evidence Units, and Scope Boundaries. This draft grounds that language in the Switchboard scenario-packet work developed in Switchboard Harness DSL: Runtime Lab Scenario Packets as Governed Agent Work Contracts by A.G. Mauro and C.A. Harris. AEL should not replace the scenario packet or runtime. Instead, it can express the authority and evidence policy inside a scenario packet so a governed run can refuse, block, or escalate for inspectable reasons. The contribution is a final-local language and methods paper: legitimate action can be declared, compiled, evaluated, and audited as a first-class contract, while legal validity and institutional authority remain outside the language unless separately supplied and accepted.