The principle of least privilege (PoLP) is a security best practice that grants each user, service, or process only the minimum access and permissions necessary to perform its legitimate function, no more. By limiting what any identity can do, least privilege shrinks the damage an attacker can cause if that identity is compromised.

The logic is containment: if every account has only the access it truly needs, then compromising one account gives an attacker limited reach, rather than the keys to the kingdom. Least privilege is a foundational concept in access control and a core pillar of zero trust.

Why least privilege matters

Over-provisioned access is a major security liability. When users accumulate permissions they don’t need (privilege creep), or when accounts are granted broad access "just in case," a single compromised credential can expose far more than necessary. Least privilege directly limits blast radius: a compromised least-privilege account can reach only its small slice of resources. It also reduces insider-threat risk and makes audits cleaner, since access maps tightly to need.

How to apply least privilege

Applying PoLP means granting minimal permissions by default and adding access only as justified; regularly reviewing and revoking unused or excessive permissions (fighting privilege creep); using just-in-time access (granting elevated permissions only for the moment they’re needed, then revoking them) rather than standing privileges; and applying it to non-human identities and AI agents too, which should be scoped to exactly their task. Fine-grained authorization is the mechanism that makes least privilege practical at the resource level.

Least privilege for machines and AI agents

Least privilege is increasingly critical for non-human identities. Service accounts, workloads, and especially AI agents often receive broad standing access that becomes a serious risk if compromised or misused. The modern approach is tightly-scoped, ephemeral authority: an agent gets precisely the permissions its current task requires, for as long as it needs them, and no more. This "just-in-time, just-enough" access is least privilege applied to the fast-growing population of machine identities, and it’s key to containing the risk they introduce.

Least privilege and zero trust

Least privilege is one of the core principles of zero trust, alongside "verify explicitly" and "assume breach." Zero trust assumes any identity could be compromised, so limiting each identity’s privileges is essential to containing the damage. Together, continuous verification and least privilege ensure that even a successful compromise reaches as little as possible, which is the essence of a resilient, breach-tolerant architecture.

Frequently asked questions

What is the principle of least privilege?

Granting each user, service, or process only the minimum access needed to do its job, no more.

Why is least privilege important?

It limits the blast radius of a compromised account, reduces insider risk, and cleans up audits by tying access to need.

What is just-in-time access?

Granting elevated permissions only for the moment they’re needed and revoking them after, instead of standing privileges.

How does least privilege apply to AI agents?

Agents should get tightly-scoped, ephemeral access to exactly their task’s needs, containing the risk if they’re compromised or misused.

What is privilege creep?

The gradual accumulation of unnecessary permissions over time, which least privilege counters through regular access reviews and revocation.

How does least privilege relate to zero trust?

It’s a core zero-trust principle (alongside "verify explicitly" and "assume breach") that contains the damage if any identity is compromised.

Related: Zero Trust · Authorization · Fine-Grained Authorization (FGA) · Non-Human Identity (NHI) · Role-Based Access Control (RBAC)