Book a Health Check

CI/CD evidence for ISO/IEC 27001

Connect a delivery risk to a configured control, a named owner and evidence that someone actually reviews.

Originally published .

Practical guidance, with sources and trade-offs.

The engineering decision

A pipeline can support an information security management system. It cannot establish organisational certification on its own. Start with a risk and show how the delivery workflow treats it.

On this page

Start with the risk

ISO/IEC 27001 specifies requirements for an information security management system. The organisation defines its scope, assesses risks and selects appropriate controls. A pipeline is one part of that system, alongside people, policies, operation and continual improvement.

For a delivery team, a useful starting risk might be an unauthorised or insufficiently checked change reaching production. Agree the risk owner and the evidence they need, then map the actual delivery path.

One control, through to evidence

If any columns are cut off, scroll sideways to compare them. Keyboard: focus the table and use the arrow keys.

Illustrative change-control example; configure and test it for your own scope.
Control intentImplementation to assessEvidence to retain
A change is independently reviewedApproval rules, eligible approvers and protection against inappropriate self-approval or rule overridesMerge request, diff, approval identity/time and relevant settings
Required checks inform the releaseDefined pipeline checks and an exception/escalation routePipeline results, findings and the decision to accept or remediate them
Production access is controlledDeployment permissions, protected environments and appropriately scoped credentialsDeployment record, actor, approval and access review

GitLab documents approval settings and protected environments separately. Their availability and effect depend on your tier, role and configuration. Test normal, exception and privileged-user paths; do not assume that displaying a rule makes it impossible to bypass.

Protect the checks themselves

Review who can change CI configuration, run privileged jobs or read secrets. Restrict credentials to the task and environment that need them. Assess runner isolation and the trust boundary for externally contributed changes.

GitLab’s pipeline-security guidance recommends using specific references for included configuration. Pin shared components, review updates and retain the version that ran with the change. That makes a check’s provenance easier to explain.

Prove it with a recent change

  1. Choose a completed production change and identify the risk/control it should demonstrate.
  2. Trace the request, diff, reviewer, pipeline and deployment records.
  3. Ask someone outside the change to explain why it was approved using those records.
  4. Record missing evidence, an owner and a remediation action.

Use the same path when AI helps draft the change. Record the intended task, permitted access and what a person checked. An AI-generated diff still needs the configured controls and an accountable acceptance decision.

Also walk an urgent-change exception. It should have an authorised decision and a review trail, rather than an undocumented route around the normal process. Evidence retention and access should match the organisation’s agreed requirements.

Turn findings into engineering work

Prioritise the gaps that weaken the decision: unclear approvals, unreliable checks, excessive credentials or missing operating ownership. Define an acceptance test for each improvement, then review it with the risk owner.

A Health Check can connect the platform configuration and delivery workflow to a prioritised roadmap. The scope, participants, timing and cost are agreed before the review begins.

Primary sources

Source scope and product terms can change. Checked on 7 November 2026.

Partnerships & assurance

Scroll to see allUse the left and right arrow keys to scroll.