A measuring device running an end-of-support operating system. An old application that doesn’t support MFA. A file transfer method imposed by a customer. Every policy sooner or later meets a situation where it can’t be applied. When these are brushed aside with “let’s open it for now” approvals by email, permanent gaps nobody owns build up within a few years: firewall rules nobody knows who opened, service accounts whose passwords never change. Below we explain how to distinguish a policy exception from a nonconformity, the request form, the risk and compensating control assessment, approval authority, time limits and tracking the exception register.
The difference between an exception and a nonconformity
Both are deviations from policy. The difference is whether the deviation was accepted knowingly, by an authorized person, with the risk assessed and for a limited time.
| Feature | Exception | Nonconformity |
|---|---|---|
| How it arises | Requested in advance | Detected in an audit, incident or check |
| Risk assessment | Done before approval | Done after detection |
| Approval | Granted according to the authority matrix | No approval; a corrective action is opened |
| Duration | End date mandatory | Closure date set by the corrective action |
| Equivalent in the standard | Risk treatment and risk acceptance (6.1.3, 8.3) | Nonconformity and corrective action (10.2) |
In an audit it plays out like this: an unapproved deviation is written up as a nonconformity. If the same deviation is covered by a recorded, valid exception, the auditor usually questions how the exception process works and whether the compensating controls are actually applied. An expired exception is a nonconformity again. How findings are classified is covered in the article major vs minor nonconformity.
There’s one more distinction: between exceptions and situations the policy already provides for. If the policy says “for service accounts, password length and restricted sign-in are applied instead of MFA”, that’s not an exception, it’s a policy rule. If the same exception is reopened every quarter, the problem isn’t the request, it’s the policy. Update the policy.
Fields of the exception request form
The form must carry the information the approver needs to decide without looking at another source. Minimum fields:
- Requester and business owner: The person opening the request and the manager who will own the risk are written separately.
- Relevant policy clause: Document code, version and clause number (e.g. “IS-POL-04 v3.1, clause 6.2 – MFA requirement”).
- Scope: Affected systems, accounts or users; with asset inventory IDs.
- Business justification: Why the policy can’t be followed, and the impact if compliance were attempted.
- Alternatives considered: Why they can’t be applied.
- Compensating controls: What will be done, who is responsible, how it will be verified.
- Residual risk level: According to the organization’s risk methodology.
- Requested end date and permanent fix plan: What the exception will end with.
The form has no “indefinite” option. A situation with no permanent fix is either added to the policy as a written rule or managed in the risk register as a risk acceptance. The gap we see most in the field is the “scope” field: “production machines” is written, but which IPs and which accounts isn’t. An exception with a vague scope ends up covering everything over time.
Risk assessment and compensating controls
Don’t set up a separate methodology for exceptions. Use the likelihood–impact scale and acceptance criteria from the ISMS risk assessment, so exception decisions stay comparable with the risk register. How to set acceptance criteria is the subject of the guide risk acceptance criteria and risk appetite.
The assessment has two stages: first the risk if the control isn’t applied at all, then the residual risk with compensating controls. A compensating control must reduce the risk the original control reduces to a similar level. Adding an unrelated control to the list doesn’t make the assessment valid; “antivirus installed” doesn’t compensate for missing MFA. A typical scenario from the field:
A test bench in production runs an operating system the manufacturer no longer supports, and the software doesn’t work on the new version. The policy requires a supported operating system and current patches. Compensating controls: moving the device to a separate network segment that allows only the necessary ports, cutting internet access, blocking USB storage access, keeping a device-specific local administrator password, and reviewing traffic logs in the segment monthly. Permanent fix: adding the test bench replacement investment to the next budget period.
Each control here narrows an exploitation route for the unpatched system (network, media, identity). The quickest way to verify segmentation is to scan the device’s ports from a client outside the segment and see that no “any” remains in the firewall rule. A VLAN on paper isn’t enough. For technical decision options for systems that can’t be patched, see the OT vulnerability management guide.
Approval authority matrix
The person approving the exception must have the authority to accept the residual risk. ISO/IEC 27001:2022 clause 6.1.3 requires the risk treatment plan and residual information security risks to be approved by risk owners; exception approval is tied to this chain of authority too. A typical authority matrix:
| Residual risk | Approver | Maximum duration (example) |
|---|---|---|
| Low | Asset/process owner + ISMS manager | 12 months |
| Intermediate | Unit director + ISMS manager | 6 months |
| High | Top management or information security committee | 3 months |
| Critical | No exception granted; risk treatment mandatory | – |
The durations in the table are example values to be set by the organization. Two principles don’t change. The requester can’t approve their own request. The ISMS manager doesn’t approve an exception alone; their job is to ensure the assessment is accurate and the record complete. The role distribution should be shown as an exception row in the ISMS RACI matrix.
Time limits and renewal
What makes an exception an exception is the end date. The duration is set by the realistic completion time of the permanent fix and the risk level, and can’t exceed the maximum in the authority matrix. An undated exception is a permanent gap.
Write the renewal rules from the start:
- A renewal request is opened as a new request before the end date; the old record isn’t extended automatically.
- At renewal, evidence is required of what was done toward the permanent fix in the previous period and that the compensating controls worked.
- The number of consecutive renewals of the same exception is limited; when the limit is exceeded, the decision moves up one approval level.
- If there’s no permanent fix on the end date and renewal wasn’t approved, the situation is recorded as a nonconformity.
“Evidence” at renewal means concrete output: an export of the segment firewall rule, the note from last month’s traffic review, a report showing the USB blocking policy is applied. The sentence “controls are continuing” isn’t evidence.
Tracking and reporting the exception register
All exceptions are kept in a single register. The register links to the approved forms and contains the fields needed for tracking:
| Field | Example value |
|---|---|
| Exception no. | EXC-2026-014 |
| Policy clause | IS-POL-04 v3.1 cl. 6.2 |
| Scope | IN-22 test bench |
| Residual risk / approver | Medium / Production Director |
| End date / status | 2027-03-31 / Active |
If the register is kept in a table, finding exceptions with 30 days left is a simple query. The PowerShell example below reads a CSV export in which the EndDate column is kept in yyyy-MM-dd format:
$limit = (Get-Date).AddDays(30)
Import-Csv -Path .\exception-register.csv |
Where-Object {
$_.Status -eq 'Active' -and
[datetime]::ParseExact($_.EndDate, 'yyyy-MM-dd', $null) -le $limit
} |
Select-Object ExceptionNo, Scope, Approver, EndDate
Running this query weekly as a scheduled task and sending the output to the ISMS manager takes the reminder out of a person’s memory. Whether the register sits in a table or a workflow tool, the regular report should include these indicators: the number of active exceptions and their distribution by risk level, those expiring in the coming period, records past their end date but not closed, those renewed more than once, and concentration by policy clause. The last indicator shows which policy clause doesn’t match reality and is a direct input to policy review.
This report is presented as input to management review; accepted risks thus come before top management. How the exception process fits into the policy set should be thought through together with the policy structure described in the information security policy writing guide. For carrying out the residual risk assessment consistently with the organization’s risk methodology, the corporate information security risk assessment page explains the relevant framework.
Rollout checklist
- Has the exception process been published as a written procedure in the policy set?
- Does the request form exclude an “indefinite” option and make a reference to the policy clause mandatory?
- Has the approval authority matrix been approved by top management?
- Have existing informal approvals (email, verbal) been recorded or closed one by one?
- Is an automatic reminder or periodic query defined for exceptions nearing expiry?
- Has the exception report been added to the management review agenda?
The fourth item takes the most time. Scanning the description fields in firewall rules and old approval emails reveals most unrecorded exceptions. To set up the exception process together with your existing policy and procedure set, see the information security policy and procedure package page.
Frequently asked questions
Is a policy exception the same as risk acceptance?
Close, but not the same. Risk acceptance is accepting a risk identified in the risk assessment without treating it. A policy exception is approving a deviation from a specific policy clause, time-limited and with compensating controls. The residual risk of the exception is assessed against the organization’s risk acceptance criteria.
Who should approve an exception?
The person or board with authority to accept the residual risk. The higher the residual risk, the higher the approval level. The requester can’t approve their own request; the ISMS manager carries out the assessment but isn’t the sole approving authority.
How long should exceptions last?
The standard doesn’t give a period. The period is set by the realistic completion time of the permanent fix and the residual risk level, and doesn’t exceed the maximum in the organization’s authority matrix. No open-ended exceptions are granted.
What happens to an exception that expires?
If renewal wasn’t approved before the end date and the permanent fix wasn’t applied, the deviation is recorded as a nonconformity and the corrective action process is started.