Skip to main content

Artificial Intelligence Product Specification (AIPS)

An open specification for qualifying, describing, deploying, governing, assuring and evolving AI Products as versioned, evidence-backed Products.

AIPS distinguishes an AI Product from a Model Product, AI-related asset, or AI-enabled Product by applying BPS Product qualification together with Product Intrinsic Nature (PIN), a distinct intelligence proposition, and direct or transitive evidence of a qualifying Artificial Intelligence Primitive.

AI Product OntologyAIPRODAIPDSAIPS-DPP

The ontology comes first

Productization and AI Product qualification are different questions. A governed downloadable LLM can be a valuable Model Product without automatically becoming an AIPS AI Product. Conversely, a symbolic reasoner can qualify as an AI Product without owning a trained ML model. AIPS therefore describes intelligence provenance, capability, effect, agency, embodiment, risk and currentness on separate axes.

AI Product Type (APT) classifies intelligence provenance as Primary or Derived. Intelligence Capability (IC), Interaction/Execution Mode (MO), Intelligence Effect (IE), Agency/Autonomy (AR), Embodiment (EM), Consumption Scope (CS), Purpose (PU), topology and R0-R4 risk remain orthogonal characteristics.

Core AIPS layers

AI Product Ontology

Defines AI Product qualification, AI Primitives, Model Product boundaries, APT and the orthogonal ontology dimensions.

Open

AIPROD

Versioned semantic description of qualification evidence, intelligence meaning, lineage, governance and Product intent.

Open

AIPDS

Mechanism-neutral realization across learned-model, symbolic, agentic, hybrid and Physical AI Products.

Open

AIPS-DPP

AI Product Trust Passport for qualification, ontology, risk, governance obligations, currentness and assurance.

Open

Lifecycle, discovery and value

PDEP / HDIP Spec C

Progresses intent through preliminary qualification, definitive AQE/APT compilation, realization, trust publication and Continuous Assurance.

Open

Marketplace

Projects verified semantics and trust while preserving the distinction between AI Product, Model Product and AI-enabled Product.

Open

COVP

Observes usage, consumer feedback, fitness-for-purpose and value realization without allowing popularity to redefine ontology.

Open

Lifecycle Closure

Proves AQE, AIPROD, AIPDS, DPP, Marketplace, COVP and PMDD tell one consistent Product-version story.

Open

Boundary rule

Uses AI ≠ AI Product. Product Family is an evidence-backed, version-specific ontological claim. Model/training evidence is conditional; symbolic, reasoning/planning, agentic and Physical AI are first-class cases.