Blog/How-to guides

Break-Glass Access: Emergency Admin That Does Not Become a Back Door

You need a way back in when the usual administrator is unreachable. How to build an emergency account that works in a crisis and is not a standing weakness.

CyberWatch AI21 November 2026 · 5 min read
A red emergency box mounted on a wall with glass front

A small architecture practice tightened up its Microsoft 365 configuration one Friday. Sensible changes: require a managed device for administrator sign-in, block access from outside the country. On Monday the only administrator tried to sign in from a personal laptop while working from a relative's house abroad, and could not. The policy was working exactly as designed. Nobody could change it, because changing it required administrator access.

That is a lockout, and it does not need an attacker. The common causes are mundane: a policy that excludes its own author, a lost phone holding the only authenticator, a single administrator who is suddenly in hospital, an identity provider outage, or an IT provider who stops answering the phone.

Break-glass access is the planned answer. The name comes from the glass-fronted box on a wall: available, obvious, and impossible to use without leaving evidence.

What it actually is

A break-glass account is an administrator account that belongs to the organization rather than to a person, is almost never used, and is deliberately kept outside the dependencies that could fail. Its purpose is narrow: get an administrator back in so the real problem can be fixed.

Three properties make it work.

  1. It is not tied to an individual. No personal name, no personal phone, no mailbox anyone reads. It survives a resignation and a lost handset.
  2. It does not depend on the thing that broke. A cloud-only account, not one synced from an on-premises directory. Exempt from the conditional access rules that could block it. Its own credential, not shared with anything else.
  3. Using it is visible. Every sign-in generates an alert to at least two people, so the account cannot quietly become somebody's convenient route around normal controls.

The third property is what separates a break-glass account from a back door. Without monitoring, you have simply created an unowned full administrator with a password in a drawer, which is worse than the problem you set out to solve.

Setting one up in Microsoft 365 or Google Workspace

The steps are similar on both platforms.

  • Create a dedicated administrator account with a neutral name. Give it the highest administrator role, because a half-privileged emergency account cannot fix a policy problem.
  • Register a strong second factor that no individual carries. A hardware security key kept in the safe is the ideal, because it is phishing-resistant and physically controlled. Where that is not possible, use a long random passphrase plus the printed recovery codes, stored separately from each other.
  • Exclude it from conditional access policies that could lock it out: device compliance requirements, location restrictions, and anything tied to a specific app. Microsoft's own guidance on emergency access accounts spells out which exclusions to apply and why.
  • Do not exclude it from having a second factor at all. This is the mistake that turns the safety net into the easiest target in your tenant.
  • Turn off password expiry for it if your platform offers that, because an expired emergency password is the single most common reason break-glass access fails.
  • Alert on every sign-in, to two or more recipients, by email and ideally by text.

Keep the same thinking for the accounts outside your office suite. Your domain registrar needs a second person who can sign in. So does your banking portal, your website host, and your password manager, where the recovery kit should be printed and stored rather than left only on a device.

Where the credentials live

The storage question is the one people get wrong, usually by putting the break-glass password in the password manager that the break-glass account exists to work around.

Work backwards from the scenario. If you are locked out of your email, can you still reach the credential? If the answer involves signing in to anything, pick something else.

What works for a small business: the passphrase and recovery codes printed, sealed in a tamper-evident envelope, and kept in a safe or a bank deposit box. For a stronger version, split the passphrase in half and give each half to a different person, so neither can use it alone. Record the date on the envelope and sign across the seal, which makes unauthorized use visible.

Alongside it, keep a one-page sheet listing what the account unlocks, who is allowed to authorize its use, and the support phone numbers for your registrar, bank and office suite provider. That sheet is useful in several emergencies besides this one. Printed recovery codes for other accounts belong in the same place, which connects to backup codes.

Agree the rules before you need them

Write down four things, briefly:

  • When it may be used. Specific situations, not "emergencies". Normal administrator access is unavailable, and waiting is causing real harm.
  • Who authorizes it. A named person and a named deputy.
  • What happens during use. Note the time, the reason and what was changed.
  • What happens afterwards. Change the credential, re-seal the envelope, restore any policy exclusions you touched, and review the log of what the account did.

The aftermath step is the one that gets skipped, and skipping it is how a one-off emergency becomes a standing shared admin login. Each use should end with the account back in its box.

Test it, or it will not work

Twice a year, open the envelope and sign in. Confirm the password still works, the security key is still registered, the licence is still assigned, and the alert actually arrives. Then change the credential, reseal and record the date.

Testing finds the failures you would otherwise discover during a crisis: a conditional access policy added in March that swept the account up, a licence removed during a cost review, an alert rule pointing at a mailbox nobody checks. If a sign-in alert for the account ever arrives unexpectedly, treat it as serious, and if the alert itself looks odd you can paste it into CyberWatch AI Scan for a free check before acting on it. Add the check to whatever rhythm you already have, such as your quarterly access review.

If you use an external IT provider, ask directly: what happens if your company disappears tomorrow, and can we still administer our own tenant and domain? The answer should be yes, and you should have tested it.

Break-glass access is insurance against the small, boring disasters that take a business offline for three days. It costs one account and one envelope. For the rest of the administrator picture, see administrator account security, least privilege, and protecting small businesses from digital threats.

Frequently asked questions

What is a break-glass account?

A seldom-used administrator account that exists so you can still get into your own systems when the normal route fails. It is deliberately kept outside the usual dependencies, and using it should be a notable event.

Why not just rely on having two administrators?

Two administrators is the right starting point and does not cover every case. Both can be locked out by the same failure: a policy misconfiguration, a lost second factor, an expired licence or an identity provider outage.

Should the break-glass account have a password or a security key?

A security key is stronger and should be preferred where the platform supports registering one for an account that no individual owns. A long random passphrase stored sealed is the fallback, with the recovery codes stored separately.

Where should the credentials be kept?

Somewhere that does not depend on the systems it unlocks. A sealed envelope in a safe, or split between two sealed envelopes held by two people, beats a password manager entry that you cannot reach because you are locked out.

How do we know if someone misuses it?

Alert on any sign-in to that account, sent to more than one person. The account should be used so rarely that any sign-in is worth a phone call, which makes the alert genuinely useful rather than noise.

Should the account be excluded from multi-factor authentication?

No. Exclude it from policies that could lock it out, such as conditional access rules tied to a device or location, but keep a strong second factor registered to it. Excluding it from MFA turns it into a password-only administrator.

How often should it be tested?

Twice a year is reasonable. Untested emergency access fails when needed, usually because the password expired, the licence lapsed, or a new policy swept it up.

Does this apply if an IT provider manages everything?

Especially then. You should hold an administrator route into your own tenant, domain and key systems that does not depend on the provider continuing to exist or continuing to like you.

Sources

  1. Manage emergency access admin accounts in Microsoft Entra ID, Microsoft Learn
  2. Security best practices for administrator accounts, Google Workspace Admin Help
  3. Identity and access management, UK National Cyber Security Centre
  4. Incident management, UK National Cyber Security Centre
  5. Implementing Phishing-Resistant MFA, US Cybersecurity and Infrastructure Security Agency
  6. Logging and monitoring, UK National Cyber Security Centre
  7. Microsoft Entra Conditional Access, Microsoft Learn
  8. Small organisations guide to cyber security, UK National Cyber Security Centre
Share this article
LinkedIn WhatsApp X
For organizations

Your people get these messages at work too.

CyberWatch AI sends your staff realistic practice attacks, trains the gaps it finds, and shows management exactly where the organization stands.