In AI Governance, Fact, Policy, Evaluation, and Process must be distinguished as different types of information. Separating them makes it possible to independently determine what existed or occurred, what Governed the Situation, how it was evaluated, and what processing the System actually performed.

A Fact Tag represents what existed or occurred based on Reality or Observed Evidence. A Fact is not altered by an Organization’s judgment, but is treated as an Evidentiary Source supporting Evaluation and Verification. Vigilance observes Facts and uses them for Evaluation and Governance Feedback, but does not rewrite the Facts themselves.

A Policy Tag represents what applies to a Fact or Case and what is required from a Governance perspective. Its Source lies in Institutional Governance, including Law, Regulation, Standards, Institutional Requirements, and Organization Policy. Organizations may have some discretion in application and contextualization, but must maintain Alignment with higher-level Governance. Where Vigilance identifies a Policy Gap or other issue, the Finding can be fed back into Governance and connected to Review through the appropriate Authority and Process.

An Evaluation Tag represents how a Fact was evaluated against Applicable Policy and Evaluation Criteria. Evaluation Frameworks, Thresholds, Sensitivity, Weight, and related parameters may be configured according to Risk, Domain, Organization, and Operational Context while maintaining Alignment with Governance. An Evaluation Tag is neither Fact nor Policy. It records the Evaluative State produced by relating them through an Evaluation Framework.

A Process Tag records the processing that the System actually performed. It traces the path actually taken through Input, Processing, Decision, Execution, and related stages, rather than a Process that can be retrospectively altered at the Organization’s discretion. Vigilance can analyze this Processing Trace to determine what Process a Decision or Execution followed and whether Anomalies or Deviations occurred.

By maintaining these four types of Tags separately, Evidence becomes more than an Event Log. It becomes a Structured and Verifiable Record. The Fact Tag functions as the Evidentiary Source, the Policy Tag as the Applicable Governance Basis, the Evaluation Tag as the Evaluative State, and the Process Tag as the Processing Trace. By associating these Records through mechanisms such as Decision-State Evidence, it becomes possible to reconstruct “what happened,” “what applied,” “how it was evaluated,” and “how it was actually processed.”

In an Accounting analogy, the Fact Tag resembles a Receipt or Source Document, the Policy Tag an Applicable Rule, the Evaluation Tag a Structured Annotation or Assessment, and the Process Tag a Processing Record or Processing Trail. The purpose of this Analogy is not to treat AI Governance as Accounting. Rather, it illustrates the structural principle of separating Records with different characteristics while maintaining their relationships so that Evaluation and Evidence remain traceable.