Definition
Semantic drift is the gradual divergence of meaning that occurs when a concept is used across multiple systems, teams, or domains without shared governance.
It is not a failure of communication. It is a natural consequence of organizational complexity.
How Semantic Drift Occurs
Organizations grow. Teams specialize. Systems multiply.
A term that was once defined informally — through shared understanding, convention, or a document that no longer circulates — begins to carry different meanings in different parts of the organization.
No single change causes semantic drift. It accumulates over time, through the ordinary operation of complex systems.
Examples
"Risk" may mean market risk to a finance team, operational risk to an audit function, and algorithmic bias risk to an AI governance team.
"Customer" may mean a policyholder to an insurance system, a billing contact to a finance system, and a data subject to a compliance team.
"Incident" may carry different definitions under DORA, ISO 31000, and an internal IT operations framework — all operating within the same organization simultaneously.
Why AI Amplifies Semantic Drift
AI systems encode meaning at training time.
When multiple AI systems are trained or prompted using concepts that carry different meanings across domains, semantic drift becomes embedded in the systems themselves.
Unlike human organizations, which can clarify meaning through conversation, AI systems cannot negotiate meaning in context. They apply the meaning they were given — consistently, at scale, and often invisibly.
This makes semantic drift in AI environments harder to detect and more consequential than in traditional information systems. See Meaning Rooms and AI.
The Relationship to Semantic Debt
Unmanaged semantic drift accumulates over time.
When the accumulated impact becomes visible — through conflicting outputs, failed audits, or compliance gaps — organizations face what can be described as semantic debt.
→ See: Semantic Debt