← All writing
PublishedRetrospective · December 20255 min readCybersecurityGovernance

Privileged access without shared administrator accounts

Part of a retrospective. The retrospective month groups the topic; it is not an earlier publication date. Published 18 September 2026.

Replace shared administration with attributable access while preserving service identities, emergency recovery and a safe transition for the support team.

A shared administrator account can look efficient because everybody knows how to use it. It also makes a simple question difficult to answer: who made this change? When several people use the same identity, an account name in a log is not a reliable record of the individual responsible.

Removing that account is not just a password change. It is a transition in how people administer systems, how unattended jobs run and how the organisation recovers when normal access fails. Treat those as separate designs or the shared account will reappear under a different name.

Find out what the account is doing

Before disabling a widely used account, identify where it signs in and what depends on it. Ask the engineers and the service provider, then compare their answers with available records. Scheduled tasks, old integrations and recovery procedures may depend on an identity originally created for a human administrator.

Separate human administration from service operation. A person needs attributable access associated with their role. An unattended process needs an appropriately managed service identity with a technical owner, necessary permissions and a supported credential lifecycle. An emergency access arrangement has a different purpose again.

ASD's guidance warns against approaches that simply minimise the number of privileged accounts or implement shared, non-attributable privileged accounts. Fewer accounts can look tidy while weakening accountability. The objective is controlled, justified privilege, not the smallest number in the directory.

Record uncertainty during discovery. An unexplained sign-in is a reason to investigate a dependency, not proof that the account can be removed safely.

Give administrators a distinct working identity

Use named administrative access separate from ordinary email and day-to-day work. Assign permissions for the tasks people perform, rather than copying the broadest existing role to every technician. Review supplier administrators with the same attention as employees.

The separation must be usable. Staff need an approved way to reach administrative tools, elevate where authorised and request access they do not routinely hold. If the new process makes ordinary support impossible, people will lobby for permanent broad access as a workaround.

ASD advises restricting privileged accounts' internet access except where explicitly required for cloud administration, and then limiting it to the task. Translate that guidance into a supportable environment rather than making an untested blanket change. Cloud administration often needs connectivity; ordinary browsing under a powerful identity does not follow from that need.

Identity security when convenience keeps winning is relevant to the enrolment and recovery side. Named accounts still need appropriate authentication, and urgent recovery should not quietly reverse the separation.

Move a dependency before retiring its old access

Consider a hypothetical engineering firm whose shared administrator identity also runs a nightly document export. Disabling it on Friday would improve the account register while breaking an operation the business expects on Monday.

A safer transition assigns the export a suitable service identity, documents its purpose and tests the minimum permissions required. The team runs a controlled validation, checks the destination and observes the next scheduled execution. Only after that evidence is available does it remove the export's dependency on the shared identity.

Human users move through a different test. Each administrator demonstrates the ordinary support tasks through their named access. A second person checks that the resulting records identify the right user and action. The team then withdraws the shared interactive access through an approved change.

Have a rollback arrangement, but do not make it an indefinite escape hatch. Define when rollback is allowed, who authorises it and what evidence is required before trying again. A failed transition should produce a specific defect, not a general conclusion that named access is impractical.

Keep emergency access genuinely exceptional

Normal authentication or approval services may be unavailable during an incident. Plan a protected recovery route appropriate to the platform, with authorised custodians and a record of use. Where a platform's emergency mechanism cannot provide ordinary individual attribution, compensate with controlled retrieval, approvals and independent records.

That narrow arrangement is different from sharing a daily administrator password in a team chat. Do not treat every difficult support ticket as an emergency or distribute recovery secrets broadly for convenience. Test the route under controlled conditions and verify that authorised people can locate the instructions without relying entirely on the service they are recovering.

After emergency use, follow the platform's supported process to restore protection, review activity and change exposed credentials where appropriate. Document the event even if the use was legitimate. An emergency account that is never reviewed can gradually become ordinary infrastructure.

Close the access lifecycle

Assign ownership of privileged roles and review whether people still need them when duties change. ASD specifically recommends revalidating privileged access regularly and when staff change duties or leave. An administrator who becomes a project manager should not retain broad access merely because it may be useful one day.

Include service identities in the review without treating them like employees. Ask whether the service still exists, whether its permissions remain appropriate and whether the credential lifecycle works. Removing a departed staff member's account does not necessarily remove an integration they created.

The detailed closure checks in how to check that offboarding really removes access help here. Success is a support team that can do its work through accountable access, with a tested emergency route and no hidden dependence on yesterday's shared password.

Working on something similar?

If this connects with something you’re working through, I’m happy to talk about where you’re stuck and whether I can help.

See how I work