Lifecycle Closure Examples
1. Current Primary Fraud Detection AI Product
The canonical v0.1 Fraud Detection example is aligned across:
- Core AI Product Version;
- AIPROD;
- AIPDS;
- AIPS-DPP;
- Marketplace listing;
- COVP observation set;
- PMDD assessment;
- PDEP lifecycle record.
All artifacts refer to Product Version urn:aip:fraud-detection:1.0.0 and agree on:
APT = Primary;IC = Detection + Classification + Prediction;AR = Recommend;EM = NonEmbodied;Risk = R2;Topology = DomainCoupled;AI Qualification = CURRENT;- current trust state for active Marketplace consumption.
The Stage 12 validator intentionally includes negative copies where the DPP points at another Product version or the Marketplace projects a different APT. Each individual artifact can still be structurally valid; the bundle fails because the Product story is inconsistent.
2. AI Primitive removal and Product Family reclassification
AI Product A v1 uses a Qwen Model Product and qualifies as a Primary AI Product. A v2 replaces that mechanism with ordinary deterministic if/else software and has no other qualifying AI Primitive.
The lifecycle consistency record captures:
- v1 as historical AI Product evidence;
- v2 AI qualification as
INVALIDfor the AI Product Family; - no APT on the non-AI current state;
- Marketplace action to historicalize/suspend the AI representation;
- downstream versions pinned to v1 as unaffected;
- downstream versions adopting v2 as requiring Stage 7 requalification.
This is a controlled reclassification, not a retroactive rewrite of v1.
3. Supersession without semantic corruption
When a Product version is superseded:
- previous DPP and Marketplace semantics remain historically discoverable where policy permits;
- active entitlements follow current policy rather than forcing obsolete trust state;
- COVP remains bound to the version actually observed;
- PMDD scores remain evidence about the assessed version;
- a new version gets its own AQE, ontology, governance and trust evidence.