Glossary
This glossary provides a shared vocabulary across the Base Product Specification (BPS),
AI Product Deployment Specification (AIPDS), and AI Product Specification (AIPROD).
It combines foundational terms (BPS) with AI-specific definitions (AIPDS, AIPROD),
ensuring clarity, consistency, and governance across specifications.
Core Concepts
Product
An entity that satisfies the true product characteristics — including deployability, self-description, discoverability, governance, and economic sustainability.
Base Product (BPS)
A meta-framework defining the core characteristics of any product, independent of its domain (data, AI, software, etc.).
Data Product
A productized form of data that satisfies BPS characteristics (deployable, discoverable, trustworthy, etc.) and is governed for consumption.
AI Product
A governed Product whose value-bearing essence is a distinct artificial-intelligence capability and whose current constitution directly or transitively contains at least one qualifying Artificial Intelligence Primitive. AI Product Family membership is version-specific and evidence-backed.
See: AI Product Definition · AIPDS · AIPROD
AI Asset
An AI-related artifact or asset used in construction, evaluation, deployment, or governance that has not itself been established as a Product. Productization does not automatically imply AI Product Family qualification.
Model Product
A Product whose value-bearing essence is a reusable learned computational capability embodied in a governed model artifact or package. A Model Product may be a Business Product for its provider without automatically being an AIPS AI Product.
Artificial Intelligence Primitive
A computational capability capable of materially contributing artificial inference, learning, perception, reasoning, generation, planning, optimization, decision formation, or intelligent action.
AI Product Type (APT)
The family-specific intrinsic classification of an AI Product by intelligence provenance: Primary or Derived. Composite is a Derived topology rather than a peer APT.
Preliminary AI Qualification
Stage 1 PDEP screening that checks whether Productness, intelligence primacy, a distinct intelligence proposition, and a plausible direct/transitive AI Primitive path exist. It is not authoritative AQE.
Definitive AI Qualification
Stage 7 PDEP compilation that resolves the exact constitution graph and issues authoritative AI Qualification Evidence (AQE), AI Product Type (APT), derivation topology, and qualification currentness.
Intelligence Behavior & Evolution Contract (IBEC)
The general PDEP Stage 6 contract for Product behavior, effects/actions, learning/evolution posture, quality/evaluation, oversight and safety. It does not assume model ownership or retraining.
Learning & Inference Contract (LIC)
Legacy/specialized compatibility term for learning/inference-oriented behavior contracts. IBEC is the general lifecycle abstraction.
Governance & Risk
Governance Metadata
Machine-readable declarations of policies, obligations, and compliance markers that apply to a product.
Effective Risk Tier
The computed AIPS R0–R4 risk posture for one AI Product version. It is supported by evidence-backed Risk Driver Assessments and is independent from APT.
Risk Driver Assessment
An evidence-backed contribution to AI Product risk, recording a driver kind, contributed tier, rationale, and optional evidence reference.
Governance Obligation Profile
The version-specific computed set of governance controls derived from Effective Risk Tier, ontology dimensions, impact/context evidence, qualification currentness, dependency currentness, and applicable policy.
Prohibited Uses
Explicit contexts or applications where a product must not be used, declared in AIPROD.
Audit Evidence
Artifacts (logs, fairness reports, compliance certificates) used to demonstrate alignment with governance obligations.
Trust & Transparency
AIPS Digital Product Passport (AIPS DPP)
The signed, version-specific trust projection of an AI Product. It exposes AI qualification, ontology posture, effective risk, governance obligations, trust currentness and assurance while treating Model Cards/training evidence as conditional specializations.
Trust Currentness
The version-specific status of qualification, dependency, governance and assurance evidence used to determine whether a passport remains suitable for current discovery/onboarding decisions.
AI Product Assurance
Evidence that the Product's governed intelligence proposition remains fit for its intended use. Assurance is mechanism-neutral and can evaluate learned, symbolic, agentic, hybrid or Physical AI Products.
Explainability
Mechanisms that make outputs of an AI Product understandable to humans (e.g., feature importance, counterfactuals).
Transparency
Declarations about training data, model architecture, governance policies, and limitations of a product.
Observability Signals
Telemetry, rationales, and fairness indicators that provide visibility into product behavior at runtime and over time.
Lineage
The record of dependencies, upstream/downstream products, and version history.
Provenance
The origin of a product, including data sources, models, and contributors involved in its creation.
Trustworthy Discovery Mechanism
A Trustworthy Discovery Mechanism is a catalog, registry, or marketplace that enables consumers (human, system, or agent) to find and evaluate AI Products with confidence in the accuracy, integrity, and governance of the information provided.
Key attributes:
- Authoritative → entries are curated or validated by accountable owners.
- Reliable → available, queryable, and technically dependable.
- Governed → compliant with organizational, regulatory, or ecosystem-level policies.
- Transparent → includes sufficient metadata for evaluation (purpose, risk, metrics).
- Interoperable → supports open standards for cross-platform discovery.
Principle: If an AI Product cannot be discovered via a trustworthy discovery mechanism, it cannot be fully trusted, governed, or reused.
Economic & Operational
Economic Model
The declared framework for pricing, cost, and resource consumption of a product (fixed, usage-based, hybrid, tokenized, etc.).
Discoverability
The ability of a product to be registered, cataloged, and found through trustworthy discovery mechanisms.
Interoperability
The ability of a product to integrate into broader ecosystems by adhering to open standards for interfaces, metadata, and data formats.
Composability
The ability of a product to combine with others (data or AI) into pipelines, workflows, or higher-order systems.
Extension Schema
The mechanism by which a product supports schema evolution, plug-ins, and domain-specific adaptations.
COVP — Consumption Observation & Value Plane
A version-bound evidence plane for usage telemetry, consumer feedback, contextual fitness-for-purpose, realized value and assurance escalation signals. COVP observes consumption; it does not assign Product Family, PIN, APT or risk tier.
Fitness for Purpose
A contextual assessment of whether a specific Product version is suitable for a specific intended use under stated conditions. Canonical conceptual outcomes are FIT, CONDITIONAL, UNFIT and INSUFFICIENT_EVIDENCE.
Value Realization
Observed evidence that Product consumption contributes to an intended outcome, recorded with an explicit attribution-confidence statement rather than assuming correlation proves causality.
Attribution Confidence
The declared confidence that a Product materially contributed to an observed value outcome. COVP uses UNKNOWN, LOW, MEDIUM and HIGH as conceptual values.
Lifecycle
Product Development and Evolution Process (PDEP)
The governed lifecycle through which Product intent is validated, qualified, compiled, realized, published, assured and evolved. AIPS Spec C specializes PDEP for AI Products in HDIP.
Continuous Assurance
PDEP Stage 11 discipline that monitors runtime behavior together with qualification, dependency, governance and assurance currentness, and routes material changes back to the earliest affected stage.
Lifecycle Stage
Declared maturity of a product (Experimental, Beta, Production, Deprecated, Retired). This is distinct from the numbered PDEP process stage.
Versioning
The controlled evolution of a product, typically using semantic versioning (MAJOR.MINOR.PATCH).
Retirement
The withdrawal of a product from active use, while preserving metadata for audit and lineage.
Discovery & Cataloging
Catalog Entry
The registration of a product in a searchable catalog or marketplace, including metadata, lineage, and governance markers.
Lifecycle Consistency Record
A version-specific evidence record proving that AQE, APT, AIPROD, AIPDS, DPP, Marketplace, COVP, PMDD and PDEP projections describe one coherent Product reality. It references authoritative artifacts; it does not replace them.
Cross-Artifact Consistency
The requirement that independently versioned AIPS artifacts agree on identity/version, qualification, APT, ontology projections, risk/governance, deployment binding and trust currentness for the Product version they describe.
Change Impact Analysis
Version-aware traversal of material dependency bindings to determine which downstream Product versions require requalification, redeployment, reclassification or retirement review after a material upstream change.
Historical Closure
The state in which a superseded or retired Product version retains immutable historical ontology and trust evidence while no longer being represented as the current consumable version.
Principle: Shared vocabulary is a form of governance — without it, even well-designed products risk misunderstanding and misuse.
Marketplace Listing
A version-specific discovery projection that combines Product semantics, trust/currentness summary, access policy and consumption-port metadata without redefining the Product's intrinsic family.
Listing Kind
The marketplace distinction between AIProduct, ModelProduct, and AIEnabledProduct. Listing Kind is not the same as Product Family and must preserve the underlying ontology.
AI-enabled Product
A Product whose primary intrinsic Product Family is not AI Product, while AI contributes materially in an enabling or constructional role. Using AI does not by itself make the Product an AI Product.
Entitlement
A governed right granted to a consumer to access one or more Product ports or packages under declared policy, contractual, licensing, or organizational conditions. Entitlement does not override Product trust/currentness state.
Consumption Port
A governed interface or package through which an entitled consumer uses a Product, such as an API, event stream, batch interface, agent protocol, embedded interface, downloadable package, or physical-control boundary.