Identity
Who or what is making the request — a person, or a non-human identity representing an application or workload.
Cloud fundamentals
Every provider answers the same question — who can do what, to which resource — with the same handful of building blocks under different names.
Four pieces show up in every provider's IAM model, whichever vendor's console you're looking at.
Who or what is making the request — a person, or a non-human identity representing an application or workload.
A named set of permissions, designed to be assumed temporarily rather than tied permanently to one identity.
The actual grant — a statement of which actions are allowed or denied on which resources.
Permissions are usually inherited down a tree (organization → project/account → resource), so a policy set high up can apply broadly without repeating it everywhere.
The mental model transfers directly between providers — only the vocabulary and console layout change.
| Concept | AWS | Azure | GCP |
|---|---|---|---|
| Human identity | IAM User | Microsoft Entra ID account | Google Account |
| Non-human (app/workload) identity | IAM Role (assumed) | Managed Identity / Service Principal | Service Account |
| Permission grant | IAM Policy (JSON, attached to user/role/group) | RBAC Role Assignment | IAM Policy Binding |
| Grouping resources for permission scope | Account / Organization | Resource Group / Subscription / Management Group | Project / Folder / Organization |
GCP's resource hierarchy (Organization → Folder → Project → Resource) is the most explicit about permission inheritance flowing downward. AWS Organizations and Azure Management Groups do the same thing conceptually, but it's easier to miss since a plain AWS account or Azure subscription can exist without ever touching the org-level tooling.
See IAM, IAM role, IAM policy, and service account in the Glossary. For the security principle behind role design, see least-privilege IAM in Production Security & Deployment Models.