Home / Guides / Every policy exception needs an approver, a compensating control and an expiry date
Guide, read 13 Nov 2026
Every policy exception needs an approver, a compensating control and an expiry date
Framework Coverage MatrixAll guides
What counts as an exception
A policy exception is a decision, made before the fact, that a named system, team or process will not meet a stated requirement for a stated time. It is not a missed task and it is not a finding. The difference matters because auditors treat them differently: an approved exception with a compensating control is a managed risk, while a missed control found at audit is a deficiency.
PCI SSC draws this line in its own wording. In its blog on PCI DSS v4.0, the Council says compensating controls are an option within the defined approach for entities with a legitimate and documented technical or business constraint that prevents them from meeting the requirement as stated. It also says Appendix B was clarified in v4.0 so that compensating controls cannot be used to retroactively address a requirement that was missed in the past.
Copy that rule into your policy: an exception is requested before the deviation starts. If the request arrives after the deviation, treat it as an incident and a finding, and log it as such.
The request, the approver and the compensating control
A request needs five things: the requirement being excepted, the scope (system, data and users), the business reason, the risk if nothing changes, and the proposed compensating control. Anything less and the approver is guessing.
The approver should be the person who owns the risk, not the person who owns the control. A control owner can recommend; a risk owner decides. For higher-scored risks, the register threshold in risk register basics tells you who signs: a director for a moderate exposure, an executive for a high one.
The compensating control has to be a real, tested control, not a promise. PCI DSS draws a second distinction here. The customized approach (requirement 12.3.2 and Appendices D and E of v4.0) is for entities that meet a requirement differently by satisfying its stated Customized Approach Objective, and it requires a targeted risk analysis. The Council states that compensating controls are not an option with the customized approach, but that an entity can use a compensating control for some system components and the customized approach for others under the same requirement. Your internal policy can borrow the idea: one exception per scope, one compensating control per exception, and a test attached to each.
| Field | Content | Rule |
|---|---|---|
| Exception ID | EX-2026-014 | Sequential, never reused |
| Requirement | Control ID and policy clause | Cite both |
| Scope | Systems, data classes, users | Named, not "all" |
| Reason | Technical or business constraint | One paragraph, dated before the deviation starts |
| Risk | Linked risk register ID and score | Score before and after the compensating control |
| Compensating control | Control ID with its own owner and test | Must be tested on the same cycle as other controls |
| Requester | Named person | Not the approver |
| Approver | Risk owner at the right level | Signature and date |
| Expiry | Calendar date, maximum 12 months | No open-ended exceptions |
| Review | Date of next check and result | Auto-flag 30 days before expiry |
Expiry and the plan to close it
An exception without an end date is a policy change nobody approved. Set the expiry when you approve. Twelve months is a ceiling I use for most; 90 days for anything touching customer data or a regulated scope. The number is mine, not a standard's, so state your own in the policy.
Attach a plan to close. NIST SP 800-53 Rev. 5 control CA-5, Plan of Action and Milestones, requires an organization to develop a plan to document the planned remediation actions to correct weaknesses or deficiencies noted during assessment, and to update it as new findings from assessments, audits and continuous monitoring arrive. Its guidance says such plans are useful for any type of organization. Control PM-4 asks for a process that reviews those plans for consistency with the organizational risk management strategy and organization-wide priorities for risk response.
Use those two ideas directly. Each exception carries milestones, an owner for each, and a due date. At expiry there are two outcomes: the exception is closed because the requirement is now met, or it is renewed with a new approval. Silent expiry is not an outcome. Renewal should need a higher approver than the first approval, so repeated exceptions become visible to leadership.
The exception register and its review
Keep all exceptions in one register, not in email threads. The register is a table with the fields above, and it is a first-class source for the board and the auditor. An assessor sampling a control will ask whether any exceptions touch it, so link each exception to its control ID.
Review it on a schedule. Monthly, the compliance lead reads every open exception, flags those within 30 days of expiry and those with a missed milestone. Quarterly, the approvers confirm that the risk score still matches the facts. Annually, the register is compared with the control set: a control with five exceptions in a year is a candidate for redesign, as covered in writing a common control framework.
A tool helps by reminding owners and by blocking an approval with no expiry date. It cannot decide that an exception is a good idea. That decision stays with the risk owner.
Watch for one pattern: the same requester raising exceptions for the same requirement across systems. It usually means the requirement does not fit how the company works, and the fix is a policy change, not a fifth exception.
Metrics to report on exceptions
Four numbers cover it. Open exceptions by risk level. Exceptions past expiry. Exceptions renewed more than once. Average age of open exceptions in days. Report these monthly to the compliance lead and quarterly to the board, as in continuous compliance metrics for the board.
Past-expiry is the one to watch. A count above zero means the review failed; investigate why before the next audit. For PCI DSS scope, remember the dates the Council gave: v4.0 was retired on 31 December 2024, so v4.0.1 is the only active version, and the new requirements became effective on 31 March 2025. Exceptions written against a v4.0 requirement number should be rechecked against v4.0.1, which the Council published on 11 June 2024 as a limited revision.
Where this shows up in the index
For a lean security team with no compliance hire, Secureframe leads at 8.38; exception handling is where such a team feels the lack of people first. Scytale is 3rd at 7.08.
Sources
- blog.pcisecuritystandards.org/pci-dss-v4-0-compensating-controls-vs-customized-approach, read 12 Nov 2026
- blog.pcisecuritystandards.org/pci-dss-v4-0-is-the-customized-approach-right-for-your-organization, read 12 Nov 2026
- blog.pcisecuritystandards.org/just-published-pci-dss-v4-0-1, read 12 Nov 2026
- raw.githubusercontent.com/usnistgov/oscal-content/main/nist.gov/SP800-53/rev5/json/NIST_SP-800-53_rev5_catalog.json, read 12 Nov 2026
- csrc.nist.gov/pubs/sp/800/53/r5/upd1/final, read 12 Nov 2026
Last reviewed 13 Nov 2026.