# Policy applicability worksheet

**Version:** 1.0
**Last updated:** 2026-10-26
**Source:** KLA Digital (kla.digital/blog/ai-agent-memory-outdated-policy)

---

## About this worksheet

Use this worksheet to decide whether a policy that an AI agent retrieved from memory applies to the
decision in front of it. It records the purpose of the retrieval, the policy's scope, authority and
effective dates, and the authoritative check made before execution.

It is an editorial proposal informed by Shen, Toyoda and Leung, *Revoked but Still Authoritative*
(arXiv 2609.08258), and by seven synthetic probes we ran against the authors' published retrieval
guard on 26 October 2026. It is a design aid. Its effectiveness has not been benchmarked.

## Worksheet

| Decision field | Record |
| --- | --- |
| Task purpose: historical explanation, planning or execution | |
| Decision timestamp and historical date requested, if any | |
| Customer or tenant, resource and governed action | |
| Policy identifier and version | |
| Issuing authority and approval record | |
| Effective start and end; withdrawal or supersession state | |
| Explicit predecessor or replacement relationship | |
| Approved exception, its scope and expiry | |
| Source documents and derived summaries referencing them | |
| Conflicting evidence and responsible policy owner | |
| Authoritative execution check and recorded outcome | |
| Historical records retained and access rules | |

## Integration order

1. Establish the task purpose.
2. Apply the authorized scope.
3. Resolve issuer, approval and effective interval.
4. Identify the applicable policy and any exception.
5. Prepare the context for the agent.
6. Evaluate the proposed action against the authoritative policy version.
7. Keep the decision references.

Treat unresolved applicability as an explicit state with an owner. This worksheet does not prescribe
legal retention periods or certify a system's authorization behaviour.