A security team attaches a service control policy to the root of its organization that denies all actions in AWS Regions other than eu-west-1 and eu-central-1. One month later, a review finds new EC2 instances running in us-east-1 that administrators launched after the SCP took effect. All of the instances are in a single account. What is the MOST likely explanation?
Choose one.
SCPs apply to member accounts only. The management account is outside their reach, so it is the one account where an organization-wide guardrail does not hold. Inside member accounts, SCPs restrict every principal, including the root user.
Only one account is affected, and the deny was attached at the organization root, so the guardrail failed in exactly one place: the management account, which SCPs never affect. The root-user explanation is the tempting inversion of the rule; SCPs do bind member-account root users. Accounts directly under the root inherit root-level SCPs like any other account. A permissions boundary cannot loosen an SCP because every guardrail layer must permit an action for it to succeed.
- Note the clue: the policy failed in a single account, not across the organization.
- Recall the SCP scope: all member accounts, including their root users, but not the management account.
- Rule out boundary or root-user bypasses: layers intersect, and member root users are bound.
- Conclude the workloads ran in the management account, and keep workloads out of it.
Exam tip: SCPs bind member-account root users but never the management account, so keep workloads out of the management account.
Designing Secure Access: IAM Best Practices, Roles, and AWS Organizations — the lesson that teaches this.