Change Impact & Reclassification
AIPS treats ontology-affecting change as a versioned graph event, not an in-place label edit.
Change-impact chain
Typical triggers include:
- AI Primitive addition, replacement or removal;
- upstream AI Product version change;
- Product Family reclassification;
- Primary ↔ Derived APT change;
- material IC/IE/AR/EM/topology change;
- risk/governance change;
- invalidation of qualification, dependency, governance or assurance currentness.
Pinned dependencies prevent historical cascade
Assume:
Qwen Model Product
↓
AI Product A v1
↓
AI Product X v7
↓
AI Product Z v5
A new A v2 that no longer contains an AI Primitive does not mutate X v7 or Z v5 while they remain explicitly bound to A v1.
If X v8 adopts A v2, HDIP must recompute X v8's qualification. If A was X's only path to an AI Primitive, X v8 may cease to qualify as an AI Product. Any Product version that then adopts X v8 must be evaluated in turn.
Impact propagates through adopted/material bindings, not through names or ancestry alone.
Re-entry rules
The earliest affected lifecycle stage should receive the change:
| Change | Typical re-entry |
|---|---|
| Business intent / intelligence proposition changes | Stage 1 |
| Ownership/authority changes | Stage 2 |
| Constitution, IC, IE, AR, EM or topology changes | Stage 3 |
| Risk drivers materially change | Stage 4 |
| Constraints / escalation change | Stage 5 |
| Behavior/evolution contract changes | Stage 6 |
| AQE/APT/governance compilation invalidated | Stage 7 |
| Semantic descriptor projection only | Stage 8 |
| Deployment realization only | Stage 9 |
| Trust publication / listing state only | Stage 10 |
| Observation/fitness/value evidence changes | Stage 11 |
Re-entry does not imply destruction of older versions. It produces new evidence and, where necessary, a new Product version.
Product identity can survive Product Family change
Product Identity, semantic proposition, Product Family, APT and Product Version are distinct concepts.
A Product named Fraud Assessment can preserve its business proposition while a new version changes from a learned-model-based AI Product to a conventional rules/software Product. AIPS preserves the historical AI classification of the old version and records the new version's reclassification rather than forcing continuity of the AI label.