Least privilege
The principle that an identity should have only the access it needs to do its job, and no more. It is a key way to shrink the damage a stolen login can cause.
Reading Time:
3 min
Least privilege is the principle that an identity, whether a person, a service account or an AI agent, should have only the access it needs to do its job, and no more. It applies to what can be reached, what can be done and for how long.
A credential is only as dangerous as the access behind it. When access is narrow, an attacker who steals a login reaches little. When it is wide, the same login can open critical systems. Least privilege reduces the blast radius of every identity, and limits how far an insider can go.
Over time people change roles and keep their old permissions, and temporary exceptions become permanent. Access accumulates until an account holds far more than its job requires. Least privilege erodes unless it is actively maintained.
Use real usage rather than job titles. Permissions that have not been used for a defined period, for example several months, are the first candidates to remove. Ask the owners of resources to confirm what is still required.
Zero trust assumes no request is trusted by default, and least privilege limits what a verified identity can do. Role-based access control is one way to implement it, by tying permissions to roles. Without regular review, roles themselves grow too broad.
Service accounts, API keys and AI agents often hold broad, long-lived access, because it is easier to give too much than to work out what is needed. They deserve the same discipline as people: an owner, a narrow scope and regular review.
Need to know limits access to information to those who require it for their work. Least privilege applies the same idea to permissions and actions, including administrative rights, system changes and the time for which access lasts. They support each other.
It depends on risk. Privileged accounts and sensitive systems deserve more frequent reviews, and a change of role should trigger a review immediately instead of waiting for the next cycle.
Look at the share of granted permissions that are actually used, the number of accounts with administrative rights, the number of dormant accounts and how quickly unused access is removed. The trend matters more than any single number.
A finance analyst needs to read reports and approve payments up to a limit. They do not need administrative rights on the finance system, access to the HR database or permission to change payment limits. If their account is phished, the attacker can see reports and little else.
No. It limits what an attacker or insider can do, but not what they do inside the access that remains. A valid account that acts within its permissions looks legitimate to the policy, so monitoring behaviour matters as much as narrowing access.
NIS2 and DORA expect access control and the ability to justify who has access to critical systems. Least privilege is the working principle behind that expectation.
Sharelock shows where access is excessive or unused, so least privilege can be applied where it reduces risk most, and monitors behaviour after login, down to individual transactions and queries on critical applications.
Explore more terms