Category /

Cloud Controls and the Deployment Pipeline

Cloud Controls and the Deployment Pipeline

Many cloud controls exist only in the document that describes them.

The policy says encryption is mandatory, storage is private by default and a second person reviews production changes. The architecture standard repeats it. The audit evidence is a screenshot taken the week before the audit. Then someone changes the console and nobody records it.

Organisations write controls for an environment that people no longer build by hand. Pipelines build it. The pipeline decides what is deployed, with which permissions, into which account. A control that does not run in the pipeline is a recommendation.

In practice, the gap shows up in three places. Drift comes first. A team defines infrastructure in code, an engineer alters it by hand during an incident, and the code never catches up. Every control assertion now describes the wrong state. Privilege comes second. Teams give pipeline service accounts broad rights because narrow ones broke a release. Those accounts then hold more access than any engineer, and they often escape review. Exceptions come third. A policy check fails a build, a team obtains a waiver, and the waiver never expires.

None of this is exotic. It happens when an organisation treats DevOps as an engineering matter and governance as a separate matter. The two functions meet once a year in an audit and find they have described different systems.

Regulation points the same way. DORA expects financial entities to manage ICT risk across the lifecycle of their systems. NIS2 expects security measures that an entity can demonstrate. Both are easier to meet when the pipeline enforces the control at the point of change and produces the evidence as it runs. Manual evidence gathering makes both harder.

The practical move is to make the pipeline the system of record for controls. Teams that do this write policy as code and evaluate it before deployment. A failed check blocks the release. Each exception is a logged decision with a named owner and an end date. Pipeline identities get only the rights they need to deploy, and security reviews them like any privileged user. A continuous comparison against the declared state catches drift, and the team handles it as an incident.

This changes the conversation between security, engineering and audit. Auditors test the mechanism directly. Engineers meet governance as a failing build they can fix. Security owns controls that actually run.

Ask whether a non-compliant change could reach production tonight. If the honest answer depends on someone noticing, the control does not exist.

Sources: None cited. General reference to Regulation (EU) 2022/2554 (DORA) and Directive (EU) 2022/2555 (NIS2).

Share the Post:

More Posts

What the AI Act Delay Changes

A deferred deadline is the riskiest thing a compliance programme can receive. On 29 June 2026, the Council of the EU gave final approval to the Digital Omnibus on AI. Obligations for standalone high-risk systems now apply from 2 December 2027. Obligations for AI embedded in regulated products apply from

Read More

The SOC Has a Shadow AI Problem, and It Is the SOC

Shadow AI governance programmes are built to find the marketing team pasting customer data into a chatbot, or the finance analyst running a model through an unapproved tool. Almost none of them are built to look at the security operations centre itself. That is the gap. The team writing the

Read More

Segmentation Was a Project; It Needed to Be a Programme

Most organisations that have “done” network segmentation completed a project. They didn’t build a capability. The architecture deck shows clean zones, defined trust boundaries, clearly labelled traffic flows between them. The environment on the ground stopped resembling that diagram roughly around the time the project closed out. This isn’t usually

Read More

Threat Intelligence Reports Are Not the Same as Threat Intelligence

Most organisations subscribing to threat intelligence feeds are subscribing to notification. A feed tells you what happened somewhere else, to someone else, on infrastructure that may or may not resemble yours. Knowing that is not the same as knowing what threatens you. The distinction sounds pedantic until you sit in

Read More

Identity Is the Perimeter, Governance Hasn’t Caught Up

The perimeter security model is functionally obsolete for most organisations. The network boundary that once defined the security boundary has dissolved, replaced by a distributed architecture in which users, services, and data operate across environments that no firewall can meaningfully contain. Identity has become the effective perimeter. Credentials determine what

Read More

The Board Asked About Cyber Risk. Nobody Had a Good Answer

Board members are asking better questions about cyber risk than they were five years ago. The answers they are receiving have not kept pace. That gap is structural, not a knowledge problem, and it produces consequences that extend well beyond uncomfortable boardroom conversations. The mismatch between what boards need to

Read More