# 36 governance terms, each with an example from a real source

GLOSSARY, read 13 Nov 2026

**The short answer.** Thirty-six governance terms, each with a plain definition and a sourced example from a vendor's documentation, a regulator or a standard. The terms cover sensitivity, access, audit and agents, and they are the words the exposure map and the guides use, read 13 Nov 2026.

- [access history](/glossary/access-history/): A record kept by the warehouse of which queries read which tables and columns.
- [access review](/glossary/access-review/): A check, by a person or a rule, of whether an access request or an existing grant should stand.
- [agent access policy](/glossary/agent-access-policy/): A rule that sets what a given AI agent may read or do, kept apart from the policies written for people.
- [AI data governance](/glossary/ai-data-governance/): The rules and records that decide which data an AI system may read, what it may be told about that data, and who can prove it afterward.
- [attribute-based access control (ABAC)](/glossary/attribute-based-access-control/): A way to grant access by comparing attributes of the person, the data and the situation, rather than by listing who may open which table.
- [audit log](/glossary/audit-log/): A time-ordered record of who did what in a system, such as a sign-in, a query or a policy change.
- [BI field exposure](/glossary/bi-field-exposure/): The presence of a sensitive source column in a field that a report, dashboard or semantic model shows.
- [classification inheritance](/glossary/classification-inheritance/): A new data field takes its sensitivity classification from a type or parent it is linked to, instead of a person labeling each field.
- [column mask](/glossary/column-mask/): A rule attached to a column that changes the value a reader sees, for example by hashing it or hiding all but a few characters.
- [consent and purpose limitation](/glossary/consent-and-purpose-limitation/): Personal data is collected for a stated reason and used only in ways that fit it.
- [copilot oversharing](/glossary/copilot-oversharing/): An AI assistant returns content a person is technically allowed to open but should never have been shown, because file or report permissions were set too wide.
- [data access governance](/glossary/data-access-governance/): The practice of deciding, approving and recording who can read which data, with the rules enforced in the systems that hold it.
- [data classification](/glossary/data-classification/): Sorting data into categories such as public, internal, personal or restricted, usually by scanning column names and values.
- [data loss prevention (DLP)](/glossary/data-loss-prevention/): Controls that detect sensitive content and warn, alert or block before it leaves its allowed place.
- [data minimization](/glossary/data-minimization/): Collect, keep and expose only the data a task needs.
- [data residency](/glossary/data-residency/): The requirement that data stay in a stated country or region, or move out only under set conditions.
- [data subject request](/glossary/data-subject-request/): A request from a person asking a company what personal data it holds about them, or asking for a copy, correction or deletion.
- [dynamic data masking](/glossary/dynamic-data-masking/): Hiding or altering a value at the moment a query runs, based on who is asking.
- [exposure path](/glossary/exposure-path/): The chain from a sensitive source column through transformations to the report, dashboard or agent answer where it can be seen.
- [label downstream inheritance](/glossary/label-downstream-inheritance/): A report or dataset picks up the sensitivity label of the source it was built from.
- [least privilege](/glossary/least-privilege/): Give each person, service or agent only the access the job needs, and nothing more.
- [need to know](/glossary/need-to-know/): The principle that having clearance is not enough: a person sees a piece of sensitive data only if their work requires it.
- [on-behalf-of authentication](/glossary/on-behalf-of-authentication/): A pattern where an agent calls a data system as itself but for a named person, so the request carries both identities.
- [PCI data](/glossary/pci-data/): Payment card data covered by the PCI DSS standard, mainly the card number and the items stored with it.
- [PHI](/glossary/phi/): Protected health information: health data tied to an identifiable person and held by a covered entity or its business associate under HIPAA.
- [PII](/glossary/pii/): Personally identifiable information: data that can point to a specific person, alone or when combined, such as a name, email or ID number.
- [purpose-based access](/glossary/purpose-based-access/): Access that depends on the reason for the request as well as the role.
- [report viewer log](/glossary/report-viewer-log/): A record that names each person who opened a given report or dashboard, and when.
- [retention policy](/glossary/retention-policy/): A rule for how long a kind of record is kept before it is deleted or archived.
- [role-based access control (RBAC)](/glossary/role-based-access-control/): Access is granted to roles, and people or services are given roles.
- [row access policy](/glossary/row-access-policy/): A policy attached to a table that decides which rows a query may return for the person running it.
- [row filter](/glossary/row-filter/): A rule that removes rows from a result based on who is asking, for example showing only one region's customers.
- [sensitive data](/glossary/sensitive-data/): Data whose exposure could harm a person or the company, such as personal, health, financial or secret business records.
- [sensitivity label](/glossary/sensitivity-label/): A tag that states how carefully an item must be handled, such as public, internal or highly confidential.
- [service principal](/glossary/service-principal/): A non-human identity that a program, pipeline or agent uses to sign in.
- [tag-based policy](/glossary/tag-based-policy/): An access rule written against a tag such as PII instead of a table name.

Last reviewed 13 Nov 2026. Canonical: https://aidatagovernancetools.com/glossary/
