Semantic drift is an established term in ontology research describing how the meaning of concepts gradually changes across ontology versions over time as a domain evolves. A term defined one way in version 1 may carry a subtly different meaning by version 5, as usage, business context, or domain understanding shifts. This is a versioning and maintenance problem: the ontology becomes documentation of what the organisation used to mean, not what it currently means. In AI systems, semantic drift creates compounding risk because AI agents treat ontology definitions as ground truth. For the broader failure mode of meaning lost between organisational intent and AI interpretation at a given point in time — regardless of versioning — see Semantic Equivocation.
