Help CD

Glossary

CALIBRE terms defined in one place.

ISO/IEC 238942023 ISO/IEC guidance on AI risk management (principles, framework, process; Annexes on objectives, risk sources, life cycle). Guidance, not certifiable; complements ISO/IEC 42001.
OECD AI PrinciplesOECD Recommendation on AI (2019, updated 2024): inclusive growth; human rights and democratic values incl. fairness and privacy; transparency and explainability; robustness, security and safety; accountability.
CoE AI Convention (CETS 225)Council of Europe Framework Convention on AI and Human Rights, Democracy and the Rule of Law. First binding international AI treaty; binds States Parties; opened for signature 5 September 2024.
UK AI principlesFive non-statutory principles from the 2023 white paper, applied by sector regulators: safety, security and robustness; transparency and explainability; fairness; accountability and governance; contestability and redress.
AICPA Trust Services CriteriaCriteria for SOC 2 examinations (security, availability, processing integrity, confidentiality, privacy), increasingly applied to AI systems.
Governance Intensity Index (GII)0-100 composite score. Combines five weighted factors (Autonomy 25%, Consequence 25%, Irreversibility 20%, Data Sensitivity 15%, Population Scale 15%). Produced by CALIBRE.
BandOne of four categorical labels applied to the GII: Low (0-25), Moderate (26-50), High (51-75), Critical (76-100). Bands drive control selection.
Consistency floorA minimum-band rule triggered by specific factor combinations. Example: autonomy>=3 AND data_sensitivity>=3 forces band >= High.
Architecture-Sensitive AmplifierA user-declared structural property of the AI system that applies its own minimum-band floor: Agentic (Moderate), Physical-Actuator (High), Cryptographic-Critical (Critical), Distributed-Ledger (High), Biometric (High).
Compositional-Coupling AmplifierAuto-set when a linked Multi-Agent Composition scores C-GII >= 40. Uplifts the individual-agent band by one level.
Trust Score0-100 portfolio-level score combining ARA residual, CALIBRE GII, and evidence coverage. Higher = better.
Runtime Authority EnvelopeCommit-time three-condition check: approval validity, scope match, state stability. Outcome is Permit, Deny, or Escalate.
Runtime Evidence ContractEvent log of seven runtime event types streamed from MLOps: quality_evaluation, access, incident, human_override, cost, drift_signal, model_version_change.
Decision ClassA category of AI decision authorised at Decision Level. Each has a named human owner and a delegation chain.
Decision EventA single occurrence of a Decision Class instance, logged for audit.
Delegation ChainThe named human trail from Board authorisation to the specific principal executing on behalf of the AI.
Jurisdictional OverlayCountry- or sector-specific extension to CALIBRE's control set. Adds controls, never lowers the base band.
Attestation ReportRegulator-ready PDF mapping CALIBRE controls to NIST AI RMF, ISO/IEC 42001, EU AI Act, MITRE ATLAS, OWASP LLM Top 10, and ISO/IEC 27001.
CALIBRE-ComposeGraph-based assessment of multi-agent AI systems. Produces C-GII 0-100 across seven compositional risk classes CC-01 to CC-07.
Composition GII (C-GII)0-100 score for a multi-agent composition, weighted mean of the seven CC-class scores.
Shadow AIAI system or agent operating inside the organisation without governance registration. CALIBRE cannot govern Shadow AI by definition.

Missing a term? See the CALIBRE paper on Zenodo.