Microsoft SC-500: Securing Break-Glass Accounts
Emergency Microsoft Entra access exists for outages and tenant lockouts, not routine convenience. Strong break-glass design maintains independent cloud-only administrative credentials, monitored custody, recurring real sign-in tests and separation from ordinary privileged workflows. The central design question is whether administrators could still recover the tenant if federation, Conditional Access or routine MFA workflows failed simultaneously. That requires separately held credentials, tested sign-in paths and alerts that make emergency use both possible and accountable.
On this page
- Model the failures that make emergency access necessary
- Create independent cloud-only identities with clear custody
- Use phishing-resistant authentication without a shared dependency
- Exclude emergency accounts from blocking policy paths
- Protect the workstation and physical credential path
- Alert on successful and failed emergency activity
- Test emergency recovery after policy changes
- Never confuse emergency use with routine privileged operations
- Keep the recovery playbook current with organizational changes
Model the failures that make emergency access necessary
Normal tenant administration can fail because a federated identity provider is unavailable, a Conditional Access rule blocks every operator, an MFA service or device is unreachable or all eligible privileged-role assignments require unavailable approvers. Emergency identities must survive the specific dependencies that ordinary administration relies on. Microsoft recommends maintaining multiple dedicated cloud-only emergency accounts with active high-impact administrative privilege under strict protection. Simply labeling a normal executive account ‘break glass’ does not remove federation, device-management or employee-lifecycle dependencies.
Walk through a scenario in which the company identity provider and administrator laptops are both unavailable. Ask which credentials can still authenticate directly to the tenant, who holds them and which secure workstation is usable. If every recovery method depends on the same phone network or password manager, the emergency plan has a common failure point. Document distinct outage assumptions and test them before accepting the design. This is a resilience exception that should be narrowly defined and visibly monitored.
Create independent cloud-only identities with clear custody
Emergency administrator accounts should not be ordinary employee identities synchronized from on-premises directories. Use the tenant’s independent cloud domain and maintain at least two supported recovery accounts so one account or physical location is not the sole route back into administration. Credential material and backup authenticators need restricted physical or procedural custody, with defined authorization to retrieve them. Role assignments must remain usable when normal PIM activation is unavailable, which is one reason these accounts are exceptional high-value identities.
Assign separate custodians and document how responsibility transfers during staff turnover. Test retrieving one credential when the primary custodian is unreachable and verify that no single employee needs to disclose a long-term secret over an insecure messaging channel. Keep account identifiers, custody records and retrieval steps available through a resilient secure process. A recovery method that can only be found after signing into the locked tenant does not satisfy its purpose.
Use phishing-resistant authentication without a shared dependency
Emergency access should use strongly protected authentication, such as supported phishing-resistant FIDO2 credentials or certificate-based methods, with a design distinct from the ordinary administrator route. The exact method and tenant support must be verified. Separate backup devices or credentials may be appropriate so losing one phone, laptop or authenticator does not disable both normal and emergency sign-in. A high-assurance setup must still be usable by the authorized responder under the outage conditions it was created to handle.
Test credential and authenticator replacement as well as ordinary sign-in. If the plan relies on one hardware security key stored in the same office as all administrator laptops, consider a different secure location. Keep strict identity verification for replacement so the recovery process does not become a weaker takeover route. Record when keys or certificates expire and how custodians are notified. Strong authentication that nobody can use during a disaster is not a practical recovery control.
Exclude emergency accounts from blocking policy paths
Conditional Access rules targeting all users or all administrators can unintentionally include emergency identities and make them unusable during a tenant lockout. Microsoft’s guidance recommends an explicit emergency group excluded from policies that might block or restrict required recovery access; report-only policies do not cause the same blocking. The exclusion should be narrow and deliberate. It does not justify disabling sign-in logs, alerting or other monitoring, and the emergency account must not become a normal way to escape an inconvenient MFA or compliance requirement.
Test a nonproduction or controlled policy change that would block an ordinary administrator and confirm that the documented emergency path still works. Inspect the applied policy results and ensure the excluded group membership cannot be changed casually. Alerts should fire on unexpected membership updates as well as sign-ins. Every new tenant-wide access rule should be reviewed against the emergency design before enforcement, because one overlooked all-users condition can invalidate years of careful recovery planning.
Protect the workstation and physical credential path
Administrative recovery may require a secure workstation independent of the compromised or unavailable employee device fleet. The team should know where a suitable managed or privileged workstation is located, how it receives necessary updates and whether it can reach the Microsoft Entra sign-in endpoint during a network outage. Physical custody, documented retrieval authorization and verified network connectivity are part of the recovery design. An emergency credential stored securely but used on an untrusted kiosk can defeat the purpose of high-assurance sign-in.
Rehearse retrieving the credential, using the designated workstation and completing a tenant administrative check while assuming the standard VPN or phone network is unavailable. Document any proxy, browser or certificate dependencies. Store the process with enough redundancy that a regional office incident does not block access entirely. After the exercise, confirm the account session is closed and credentials are returned to secured custody. The test should prove operational readiness, not just cryptographic strength.
Alert on successful and failed emergency activity
Emergency accounts should be nearly inactive under normal conditions. Any sign-in, repeated failure, credential change, authentication-method registration, role assignment or group membership change is a high-value event requiring review. Configure reporting so the relevant responder receives actionable context rather than a low-priority message buried in routine administrative noise. Audit retention should be sufficient to reconstruct a use, and investigation roles should not depend on the compromised emergency account itself.
Perform an approved test sign-in and confirm that the alert reaches its owner with the correct account and time. Next simulate a failed attempt and check whether monitoring distinguishes it from the successful validation. Document the expected escalation if the use was not scheduled. A system where the emergency identity works but its use goes unnoticed creates a powerful unmonitored privilege path. Monitoring, custody and sign-in controls must be tested as one design.
After the test, confirm there is no unauthorized session, unexpected registered method, role change or new device trust associated with the account. If alerting is broken, treat the recovery path as degraded even if authentication succeeded. The goal is a controlled exception that remains both accessible during a genuine lockout and unmistakably visible whenever somebody attempts to use it.
Validate recovery at least quarterly and after policy changes
Microsoft’s guidance recommends recurring tests, for example every quarter, and testing after meaningful Conditional Access or identity-platform changes. A spreadsheet assertion that the account exists is not enough. Validate actual sign-in, authentication method, privileged role, policy exclusion, workstation readiness and alerting. Time-bound evidence allows the security team to identify exactly which component failed. A minor change in sign-in requirements can make a previously valid recovery account unusable without changing its name or role.
Use a written exercise with date, authorized testers, expected conditions and outcome for each recovery path. Include a scenario where one account or authenticator is unavailable, then record how the second account preserves access. Fix failures immediately and repeat the complete test, not only the single failed screen. A dormant identity can only be trusted when its unusual operating conditions are rehearsed; otherwise the first real use may occur during a crisis with no margin to diagnose the problem.
Never confuse emergency use with routine privileged operations
Frequent emergency-account use indicates that normal PIM, help-desk or authentication workflows need repair. The account should not be used for convenience during change windows or to avoid ordinary approval requirements. Limit which incidents justify retrieval, name who authorizes it and record the business reason. After use, review the audit timeline and rotate or replace credential material where required by policy and the specific event. An emergency process should restore normal administration, not establish a parallel permanently active workflow.
Hold a post-use review after every real or test invocation. Verify no unexpected role, credential, policy or session changes remain and that alerts are still configured. Identify whether the normal administrator recovery design could be improved so the same failure is less likely to recur. If the only way to maintain the tenant involves using break-glass credentials weekly, redesign the routine privilege lifecycle. Exception rarity is a practical protection control.
Keep the recovery playbook current with organizational changes
Staff departures, office relocation, new device providers, tenant reconfiguration and security policy updates can silently invalidate custody and access arrangements. The emergency system needs a named owner for both the accounts and the paper or secure offline operating process. Keep custody records, authorized personnel, alternate locations, credential replacement and incident contacts synchronized with current responsibility. Do not let ordinary inactivity cleanup or HR-driven offboarding automation disable identities whose dormancy is intentional.
During a periodic review, check whether any custodian is no longer employed, whether backup authenticators still function and whether a new tenant-wide Conditional Access rule changed the recovery pathway. Test one dependency failure deliberately. The goal is not to maximize the number of permanent Global Administrator accounts; it is to keep a very small, independently accessible and consistently monitored set of recovery credentials that can restore the tenant when standard controls fail.