Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
Service control policies set the maximum permissions available to accounts in an AWS organisation. They never grant access; they define the ceiling that no local administrator can raise.

What Are AWS Service Control Policies?
A service control policy (SCP) is an organisation-level guardrail in AWS that limits what the accounts beneath it are allowed to do. It grants nothing. Its only job is to define a ceiling, so that even an account administrator with AdministratorAccess cannot step outside the boundary the organisation has drawn. SCPs are attached in AWS Organizations to the root, to an organisational unit, or to an individual account, and they apply to every principal in that account except the account root user of the management account.
Most cloud estates grow by delegation. A platform team hands a product team an account and expects them to manage their own roles. That works until someone disables CloudTrail to reduce noise, creates an IAM user with long-lived keys for a CI job, or spins up resources in a region nobody monitors. Writing ever-longer identity policies to prevent this does not scale, because the account administrator can rewrite them.
An SCP cannot be rewritten locally. Combined with IAM policy evaluation rules, where an explicit deny always wins, it gives you a small number of controls that hold regardless of what happens inside the account.
cloudtrail:StopLogging, cloudtrail:DeleteTrail, and changes to the central log bucket and its object lock settings.organizations:LeaveOrganization and deny modification of the IAM roles your security tooling and break-glass process depend on.aws:MultiFactorAuthPresent.Every organisation starts with the default FullAWSAccess policy attached, which permits everything and leaves the ceiling wide open. From there you have two design choices. A deny list keeps FullAWSAccess in place and layers specific denies on top; it is easy to adopt and rarely breaks anything unexpectedly. An allow list removes FullAWSAccess and enumerates the services the organisational unit may use, which is far stricter and far more likely to break a deployment at an awkward moment.
Most organisations run deny lists in production and reserve allow lists for sandbox or highly regulated organisational units where the service catalogue is genuinely short. Whichever you pick, remember that SCPs are inherited and intersected down the tree: a permission must be allowed at every level to survive.
SCPs do not apply to the management account, which is one reason to keep that account empty of workloads. They do not affect service-linked roles, so AWS services can still perform their own housekeeping. They do not apply to resource-based policies evaluated for principals outside the organisation, and they have a size limit of 5,120 characters per policy, which forces careful use of wildcards and condition keys rather than long literal lists.
They are also blunt. An SCP sees the API call, not the business context, so a deny that is correct for ninety-nine teams can stop the hundredth from deploying at 2 am. Tag-based conditions and a documented exception path matter more than the policy text itself.
Attach new policies to a test organisational unit first, then to non-production, and watch CloudTrail for AccessDenied events attributed to the organisation policy before you move to production. Keep SCPs in version control alongside the rest of your infrastructure as code, review them when a new organisational unit appears, and test your break-glass role against the live policy set at least twice a year. A guardrail that also blocks your incident responders is worse than no guardrail.
Guardrails look complete on paper far more often than they are in practice. We test AWS organisation boundaries, role trust, and escalation paths as part of our cloud security services.
No. Service control policies have no effect on the management account of an AWS organisation, even when attached to the root. That account keeps full access by design, which is exactly why it should hold no workloads, no application roles, and as few human users as possible. Treat it as a control plane, protect it with hardware multi-factor authentication, and monitor every action in it.
It limits them. A stolen administrator role in a member account still cannot act outside the ceiling, so region restrictions, audit-trail protections, and denies on disabling security services continue to apply. That buys detection time and preserves evidence. It does not stop the attacker from abusing everything inside the ceiling, so SCPs complement detection rather than replacing it.
AWS allows five SCPs per root, organisational unit, or account, and each policy is limited to 5,120 characters. Because policies inherit down the tree, the effective limit is reached sooner than teams expect. Group related controls into themed policies, use condition keys and wildcards instead of long resource lists, and prune policies that duplicate controls already enforced higher up.
They are useful evidence of preventive control, and auditors generally accept them alongside the attachment map showing which accounts are in scope. On their own they prove intent rather than outcome, so pair them with AWS Config rules, Security Hub standards, and records of exceptions granted. Evidence is stronger when you can show both the guardrail and the monitoring that confirms it held.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.