in this guide
Somebody opened your company’s cloud account. They typed an email address, chose a password, added a card and, if prompted, scanned a code with their phone. That sign-up created an identity that sits above every admin role you have set up since.
Each of the three big clouds has one. In AWS it is the root user. In Microsoft it is the Global Administrator role the tenant creator received, which can reach into every Azure subscription and into billing. In Google Cloud it is the Workspace or Cloud Identity super admin, who stands over the organisation and its administrators. You can tighten every role below it. It can grant them back.
This guide is about that one account: what it can do in each cloud, what it hangs on, and how to move it off a person. The founder departure guide asks whose name is on it. The break-glass guide covers running emergency admin accounts in Microsoft Entra. This one sits between them, at the top of the tree.
one account at the top of each cloud
The vendors’ own documentation describes the top account in different words. Read side by side, the shape is the same.
AWS: the root user. AWS’s IAM documentation says the email address and password used to create the account are the root user’s credentials, and that this identity has complete access to all services and resources in the account. Its list of tasks only the root user can do includes, for a standalone account, changing the root email address and password, closing the account and reopening it. It also includes restoring permissions when the only IAM administrator has locked themselves out, and switching on IAM access to the billing console. The same page notes that account name, contact details and alternate contacts do not need root.
Microsoft: the first Global Administrator. Microsoft’s quickstart for creating a tenant says the user who creates it is automatically assigned Global Administrator and listed as the technical contact. That role reaches past the directory. Microsoft’s page on elevating access says a Global Administrator can switch one setting and receive the User Access Administrator role at root scope, which lets them assign roles in every Azure subscription and management group in the tenant. Microsoft’s billing documentation adds: “The Global Administrator role is above the Billing Account Administrator”, and says Global Administrators can add themselves as billing administrators at any time.
Google Cloud: the super admin. Google’s documentation on organisation resources says one is created for you once a Workspace or Cloud Identity account exists and someone creates a project or billing account, and the super admin who creates it is assigned Organization Administrator. Google lists the super admin’s jobs as assigning that role, being the point of contact for recovery issues and controlling the lifecycle of the account and the organisation. The same page says every user in the domain gets Project Creator and Billing Account Creator at the start.
how the top account reaches everything, in Microsoft's documented path
- 1tenant createdThe person who signs up becomes Global Administrator.
- 2one toggleAccess management for Azure resources set to Yes.
- 3root scopeUser Access Administrator over every subscription and management group.
- 4any Azure role, anywhereBilling is a separate step: Global Administrators can add themselves as billing administrators.
Below the top account you can remove anything. The top account can put it back.
what it hangs on
The top account is created by one person on one day, so it inherits that person’s details. The vendor pages show which of those details decide who can get back in.
what the top account depends on
- the sign-up email inboxpassword resets and vendor notices go here
- the phone number on the accounta recovery channel when every MFA device is gone
- the MFA devicean authenticator on a phone that may have been replaced
- where the password livesa personal manager, a shared vault, a tool inside the same cloud
- the billing cardwhose card, and who sees the invoices
decides who can recover the accountdecides who notices
The inbox. AWS’s troubleshooting page says that to reset a lost root password you must know the email address used to create the account and have access to that inbox. If nobody is sure which address that was, AWS suggests searching your organisation’s mail for messages from @signin.aws or @verify.signin.aws. If the address is on your own domain and the mailbox was deleted, AWS suggests restoring it or setting up a catch-all. If the address is a personal one, none of those options is yours to use.
The phone. AWS’s best-practice page says that if every MFA device on the root user is lost, recovery needs both the email and the phone number used to register the account, up to date and reachable. The same list of root-only tasks says that when a key in AWS’s key management service becomes unmanageable, Support confirms the request through the root user’s primary phone number. A number that belonged to someone who left is a dead end on both paths.
The MFA device. AWS requires MFA on the root user for every account type, and its best-practice page says it must be registered within 14 days of the first sign-in attempt. The dates for that change and its Microsoft and Google equivalents are in the vendor login changes guide. A requirement met on day one with the founder’s authenticator app still depends on that phone. AWS allows up to eight MFA devices on the root user and recommends registering more than one.
The password’s home. AWS warns against keeping the root password in a tool that itself depends on the same AWS account, because losing the password then locks you out of the tool too. The shared vaults guide covers another home, a shared vault many people can open.
what someone does with it
An attacker who reaches the top account can use it to create more ways back in, before anyone notices the first one.
Code Spaces, a code-hosting company, closed in June 2014 after an attacker gained access to its Amazon EC2 control panel and demanded money. In the company’s statement, as reproduced by IT Writing on 18 June 2014 (The Register quoted other parts of it the same day), Code Spaces said it changed the passwords to take the panel back, but the intruder had already created several backup logins. The attacker then removed snapshots, storage buckets and machine images, and the company said most of its data and backups were partly or fully deleted. The sources do not say whether the attacker held the root user or another admin login. They do show what any account at that level allows: creating more accounts, faster than the owner can change one password.
Microsoft’s own investigation of a group it tracks as Storm-0501, published on 27 August 2025, shows the Microsoft version of the same climb. The attackers found a synced, non-human account holding Global Administrator with no MFA method registered. Microsoft says they then invoked the elevate access operation, received User Access Administrator over all of the organisation’s Azure subscriptions, and assigned themselves Owner. What followed, including the deletion of backups, is told in the backup admin guide. The step that matters here is the elevation: one directory role became ownership of every subscription in two operations. The identity you can’t offboard explains how a non-human account ends up holding a role like that.
Both cases sit at the outer edge of your blast radius. Every other access review you run assumes this account is under control.
moving it off a person
Picture a 70-person agency. A co-founder opened the AWS account in 2019 with their own work address, put the company card on it and turned on MFA with an app on their phone. They have since moved to a part-time advisory role. The Microsoft tenant was set up by the first office manager, still the only Global Administrator besides the outside IT provider. Nobody at the agency has signed in as either top account this year.
Nothing there is broken. Every piece of it depends on two people staying reachable and keeping the same phones. The vendor guidance turns into a short set of moves.
Give the AWS root user a company address that reaches a group. AWS’s best-practice page recommends an email address managed by the business that forwards to a group of people, so that AWS can reach the owner even when individuals are “on vacation, out sick, or have left the business”. It also says the address should not be used for anything else. For a standalone account, changing it is one of the root-only tasks, so it has to be done by signing in as root.
Split the recovery channels. AWS says no one person should have access to both the root email inbox and the root phone number, and suggests two groups: one holding the password, one holding the MFA device, with one member of each needed to sign in. In a company of 70, two named people in different roles can be enough.
Register more than one MFA device. Two hardware security keys, kept in different places, survive one being lost. The break-glass guide covers storage and testing; the same rules apply here.
Retire root where you can. AWS announced on 15 November 2024 that organisations using AWS Organizations can centrally remove root credentials from member accounts, and its documentation says new member accounts created in an organisation have no root credentials by default. That leaves one root user to protect, on the management account, and the member accounts recover through it.
In Microsoft, separate the creator from the role. Microsoft’s quickstart recommends two cloud-only emergency access accounts permanently holding Global Administrator and not assigned to individuals; the break-glass guide covers building them. Then move the technical contact to a role address, and check that the billing account has more than one owner on the company side.
In Google Cloud, keep the super admin out of daily work. Google’s best-practice page asks for a separate super admin account with its own login, kept apart from the group that holds Organization Administrator, protected by a security key, with the first super admin’s key kept somewhere safe, preferably at your physical location. It also recommends keeping super admin self-recovery switched off. Recovering a Workspace super admin when the person is gone runs through DNS, as the founder departure guide explains.
using it rarely and watching every use
The top account should be signed into a few times a year, for the tasks only it can do. That makes every sign-in worth knowing about.
- AWS recommends alerting on root user sign-in and usage, through CloudWatch or EventBridge rules or its threat detection service, and having a written procedure for whoever receives the alert. It also recommends never creating access keys for the root user.
- Microsoft records each use of elevated access in the Entra audit logs (a preview feature, per its page) and the Azure activity log, and shows a banner when users still hold elevated access. Its guidance is to remove elevated access once the change is made.
- Google Cloud suggests an alert whenever anyone changes an allow policy through
SetIamPolicy(), which catches the moment someone grants a new role at the organisation level.
The question for each alert is who reads it. If the answer is the inbox of the person who opened the account, the alert and the risk share an address.
If the person who opened your cloud account stopped answering their phone tomorrow, which of your recovery paths would still work?
sources
- AWS IAM User Guide: AWS account root user (tasks that require root user credentials, centralized root access)
- AWS IAM User Guide: Root user best practices for your AWS account
- AWS IAM User Guide: Troubleshoot issues with the root user
- AWS What's New: Centrally manage root access in AWS Identity and Access Management (15 November 2024)
- Microsoft Learn: Quickstart, access and create a new tenant (Microsoft Entra)
- Microsoft Learn: Elevate access to manage all Azure subscriptions and management groups
- Microsoft Learn: Billing roles for Microsoft Customer Agreements
- Microsoft Threat Intelligence: Storm-0501's evolving techniques lead to cloud-based ransomware (27 August 2025)
- Google Cloud: Creating and managing organization resources
- Google Cloud: Super administrator account best practices
- The Register: Code Spaces goes titsup FOREVER after attacker NUKES its Amazon-hosted data (18 June 2014)
- IT Writing: Resilience is not backup, how Codespaces.com lost its data and its business (18 June 2014)
