A Process Tag records the Processing that an AI System actually performed. While a Fact Tag preserves as Evidence Facts obtained or observed from Reality or an External Origin, a Process Tag preserves the Processing that actually occurred within the System as Evidence. Process therefore represents the Processing Path actually taken, rather than an intended Workflow.

A Process Tag records the Processing Context at the time of processing. This may include the Model / Version or Agent, System Environment, Input, Retrieved Information, Tool / Function, Context, and Intermediate State. This makes it possible to compare Processing Conditions when the same or similar Inputs produce different results.

It also preserves a Processing Trace, including which Processing Steps were traversed, which Tools / Functions were invoked, and how Data and States transitioned. Tool Failures, Errors, Exceptions, Retries, and Unexpected Transitions are also recorded as part of the Actual Processing.

A Process Tag also traces how Processing was connected to Decision and Execution. It enables reconstruction of which Processing State led to a Decision, which Authority and Execution Conditions were associated with it, and how it was connected through the Execution Boundary to an Execution State and Outcome.

However, the Process Tag does not itself form a Decision or perform Execution. A Decision is formed through the Governance Process, while Execution occurs through Verification / Enforcement at the Execution Boundary. The Process Tag serves as an Evidence Layer that associates these elements within the Actual Processing Sequence.

What is recorded, therefore, is not “what the System was supposed to do,” but “what Processing actually occurred and what happened as a result.” Deviations from the Expected Workflow or Applicable Policy are verified by comparison with the Actual Process, and the differences themselves become Evidence for Vigilance and Audit / Assurance.

This is particularly important for AI Vigilance. Even where the Outcome itself appears acceptable, the Processing Trace may reveal Repeated Failures, Unexpected Tool Use, Authority-related Deviations, or other Signals of future Risk. The Processing Path leading to the Outcome, rather than the Outcome alone, therefore becomes an Observation Point.

Fact Tags and Process Tags both preserve Evidence of “what happened,” but from different Observation Points.

Fact Tag = Evidence of What Is / What OccurredProcess Tag = Evidence of What the System Actually Did

While a Fact Tag is grounded in Reality or an External Origin, a Process Tag is grounded in Actual Processing within the System. By separating and associating the two, it becomes possible to independently verify what occurred externally, what the System actually did in response, and the relationship between them.