These ten terms form the working vocabulary of Semantic Governance. They are designed to be cited, linked, and reused. Each term is also defined as a JSON-LD DefinedTerm on this page so AI systems and search engines can index them as structured terminology.

Semantic Governance

The discipline of defining, approving, and maintaining the meaning of business terms so every human, system, and AI agent interprets them the same way.

Semantic Governance treats meaning as enterprise infrastructure. It establishes who owns a term, how its definition is approved, how that definition evolves, and how reuse, drift, and contradiction are detected over time. Unlike data governance — which focuses on data quality, lineage, and access — semantic governance focuses on the interpretation layer that sits above data.

Example. When the Claims, Underwriting, and Reinsurance departments each use the word "Coverage," Semantic Governance ensures their distinct meanings are recorded as separate canonical concepts within the same Meaning Space, rather than collapsed into a single ambiguous definition.

Semantic Drift

The unmanaged divergence of a term's meaning across teams, systems, time, or AI models.

Semantic Drift is the silent failure mode of enterprise knowledge. It accumulates whenever a definition is restated, translated, summarized, or interpreted without governance. Left unmanaged, drift causes contradictions in reports, breaks integrations, and produces incoherent AI behavior even when underlying data is correct.

Example. In 2024 a definition of "Customer" is approved as "any party with an active contract." By 2026 marketing uses it for anyone in the CRM, finance uses it for billed parties only, and an AI agent has learned both — producing conflicting answers to the same question.

Meaning Space

A bounded interpretive context within which a term carries a specific, governed meaning.

A Meaning Space is the unit of contextual scope in Semantic Governance. The same surface term can exist as several distinct Canonical Concepts inside one Meaning Space (e.g. Insurance) when it carries different meanings in different sub-contexts (Claims vs Reinsurance). Meaning Spaces make ambiguity addressable instead of invisible.

Example. The Insurance Meaning Space contains five distinct canonical concepts for "Coverage": Policy Scope, Claims Limit, Reinsurance Cession, Product Offering, and Regulatory Control Coverage.

Canonical Concept

The single approved, versioned, owner-accountable meaning of a term within one Meaning Space.

A Canonical Concept binds together: a context-disambiguated name (e.g. Coverage (Claims Limit)), a canonical definition, a governance status (proposed, reviewed, approved, deprecated), a confidence score, and a provenance trail. It is the unit that humans approve, systems resolve to, and AI agents cite.

Example. "Coverage (Reinsurance Cession)" — the share of risk transferred from the primary insurer to a reinsurer under a treaty — is one Canonical Concept; "Coverage (Claims Limit)" is a different one.

Provenance

The recorded source, author, and context of every definition that contributes to a Canonical Concept.

Provenance answers: where did this meaning come from? Each ingestion records the raw term, raw definition, submitting department, source document, industry, resolver source (deterministic, context rule, AI, or organizational memory), resolver confidence, and context label. Provenance turns canonical definitions from claims into evidence.

Example. The Canonical Concept "Coverage (Claims Limit)" carries provenance entries from three claims-handling manuals, two regulatory filings, and one internal training document — each with author, date, and confidence.

Governance Decision

An explicit, recorded act by a governance authority that approves, rejects, modifies, or deprecates a Canonical Concept.

Governance Decisions are first-class events. They name the actor, the prior state, the new state, and the rationale. Decisions accumulate into a concept's history and form the audit trail required for regulated industries and AI accountability frameworks.

Example. After a drift event between two definitions of "Coverage (Policy Scope)", the governance authority approves the Underwriting version and deprecates the contradictory submission. Both the decision and the rationale are recorded immutably.

Organizational Memory

The append-only history of every governance event, definition change, and learning signal an organization has accumulated about its terminology.

Organizational Memory is what allows an enterprise to answer, years later, how a concept evolved, why it changed, who approved it, and how often it has been reused. It is stored as an immutable event log — not editable records — so the history can never be silently rewritten.

Example. Five years after first ingestion, "Coverage (Reinsurance Cession)" can show its first-seen date, every drift event, every governance decision, and how its confidence evolved from 0.84 to 0.96.

Concept Reuse

The act of resolving a new definition to an existing Canonical Concept instead of creating a new one.

Reuse is the primary learning signal in Semantic Governance. Each reuse strengthens the confidence of the canonical concept, reduces duplication, and builds Organizational Memory. A high reuse rate indicates stable, well-governed meaning; a low rate indicates fragmentation or insufficient context resolution.

Example. A second upload of a Reinsurance definition resolves to the existing "Coverage (Reinsurance Cession)" concept, raising confidence from 0.95 to 0.97 with no duplicate created.

Semantic Stability

A measure of how consistently a Canonical Concept holds its meaning across ingestions, reviews, and time.

Semantic Stability is derived from the concept's history: ratio of reuses to drift events, confidence trend, frequency of manual review, and time since last contradiction. It is not a marketing score — it is computable evidence that a term is governed well.

Example. "Coverage (Claims Limit)" with 43 reuses, 2 drift events resolved within a week, and rising confidence has higher Semantic Stability than a term with 3 reuses and 5 unresolved drift events.

AI Alignment (Semantic)

The state in which AI agents resolve business terms to the same governed meaning that humans and systems use.

Semantic AI Alignment is a prerequisite for trustworthy enterprise AI. Without it, large language models infer meaning from their training distribution, not from the organization. Semantic Governance provides the missing layer: a canonical, resolvable, evidence-backed meaning per term that an AI agent can cite instead of guessing.

Example. An enterprise assistant asked "what is our reinsurance coverage exposure?" resolves "coverage" to "Coverage (Reinsurance Cession)" via the canonical registry, instead of averaging a generic web-trained meaning.