AIPS Digital Product Passport (DPP) Overview
Purpose
The AIPS Digital Product Passport (AIPS DPP) is the AIPS specialization of the BPS DPP Core for immutable, versioned AI Products.
The passport is a trust projection of the AI Product, not a Model Passport. Its primary job is to let consumers, marketplaces, governance services and auditors answer:
- does this Product currently qualify as an AI Product?
- what is its APT and current ontology posture?
- what intelligence capabilities, agency, effects and embodiment does it expose?
- what is its effective R0-R4 risk posture and why?
- which governance obligations are currently in force?
- are qualification, dependencies, governance and assurance evidence current?
- what evidence supports trust in this specific Product version?
Model Cards, training-data manifests and training-pipeline evidence remain useful conditional evidence when the Product actually depends on learned models or learning processes. They are not universal AI Product requirements.
Relationship to BPS DPP Core
BPS DPP Core provides universal passport identity, issuance, integrity, lifecycle and Inline/ByRef mechanics. AIPS adds AI-specific trust semantics.
| BPS DPP Core | AIPS DPP specialization |
|---|---|
| Product subject and issuer | immutable AI Product version |
| hash/signature | signed AI trust projection |
| schema/lineage/provenance/quality/policy | qualification, ontology, risk, governance, currentness and assurance |
| Inline/ByRef | Lite summaries + Full evidence references |
Mandatory trust summaries
Every current AIPS DPP contains:
- AI Qualification — evidence that artificial intelligence is the Product's value-bearing essence and that a qualifying AI Primitive exists in the direct/transitive constitution.
- AI Ontology Summary — APT plus relevant IC/CS/PU/MO/IE/AR/EM/topology classifications.
- Effective Risk Tier — R0-R4 with evidence-backed risk drivers.
- Governance Obligation Profile — obligations derived from the actual ontology/risk posture.
- Trust Currentness — currentness of qualification, dependencies, governance and assurance.
- AI Product Assurance — product-level evidence that the intelligence proposition remains fit for its intended governed use.
Conditional evidence
| Evidence | Applicability |
|---|---|
| Model Card | when a learned model is a qualifying primitive/dependency |
| Training Data | when training/learning provenance is applicable |
| Training Pipeline | when the Product owns or materially relies on training/retraining activity |
| Risk Register | where human-readable risk evidence is required |
| Detailed assurance reports | where Lite summary is insufficient |
A symbolic reasoning AI Product can therefore be fully AIPS-DPP conformant without a Model Card or training evidence.
Lite and Full views
Lite is a compact trust summary suitable for discovery and pre-onboarding. Full retains the same mandatory inline trust summaries but adds cryptographically bound ByRef evidence for deeper audit, assurance and due diligence.
Example
A Fraud Detection AI Product may expose a Lite passport containing its AI qualification, APT=Primary, Detection/Classification capabilities, R2 risk drivers, governance obligations, currentness and assurance summary. Because its qualifying primitive is a learned model, it also exposes Model Card evidence.
A symbolic Policy Reasoner exposes the same trust-layer sections but does not require a Model Card simply to be an AI Product.
Principle
AIPS DPP proves trust in an AI Product version. It does not assume that every AI Product is a model, owns training data, or retrains.