Reference
Governed AI Glossary
Canonical definitions for the terminology used across the Keon governance standard.
Governed AI
An architectural model in which artificial intelligence systems operate under enforced policy boundaries and produce cryptographically verifiable evidence of both execution and human-facing expression.
Governance Substrate
The structural layer that enforces policy, produces receipts, and verifies artifacts. It governs without executing. The governance substrate is architecturally separated from the execution engine.
Execution Engine
The component that performs workflows under constraints it did not define and cannot modify. It operates within boundaries established by the governance substrate.
Evidence Pack
A sealed, cryptographically signed collection of execution receipts, deterministic manifests, and policy records that provides independently verifiable proof of governed execution.
Execution Receipt
A structured record generated during execution — not after — that binds policy version, execution ID, expression hash, tenant scope, timestamp, and digital signature.
Behavioral Sovereignty
The principle that any AI system presenting human-facing expression must comply with an approved behavioral policy prior to exposure. Expression is contractual, not cosmetic.
Behavioral Policy
A formal specification defining archetype declaration, lexical constraints, structural framing rules, emotional temperature bounds, agency preservation standards, and enforcement severity classifications.
Behavioral Evaluation Gate
The enforcement checkpoint through which all human-facing expression must pass before exposure. It evaluates compliance status, rewrite eligibility, violation severity, and final disposition.
Behavioral Compliance Receipt
A structured record proving that human-facing expression passed the behavioral evaluation gate with an APPROVED disposition under a specific policy version.
Archetype
A declared behavioral identity that defines how an AI system presents itself to humans. Archetypes are governed by behavioral policy and cannot drift without explicit ratification.
Deterministic Manifest
A canonical, hash-bound artifact record that ensures reproducibility and immutability. Once sealed, a deterministic manifest cannot be altered without detection.
Cryptographic Binding
The process of sealing execution artifacts with cryptographic signatures to ensure tamper-evident integrity and enable independent verification.
Separation of Powers (AI Context)
The architectural requirement that authority and action be separated. The execution engine performs; the governance substrate enforces. Neither may assume the other's role.
Audit-Ready AI
An AI system that produces sufficient deterministic artifacts, execution receipts, and cryptographic evidence to satisfy independent audit requirements without reconstruction.
Behavioral Drift
Unauthorized or untracked change in an AI system's behavioral expression. Under the Keon Canon, behavioral drift requires version increment, canon update, and explicit ratification. Silent drift is prohibited.