Principle of least privilege
Also called Least Privilege
Give every user, process, or agent exactly the access it needs to do its job — and nothing more.
Think of it like
A hotel keycard that opens your room and the gym, but not every other guest’s door.
Example
An AI agent that only needs to read a calendar is given a read-only, calendar-scoped token instead of full account access.
How it actually works
Least privilege limits the blast radius of any compromise or mistake: if a component is breached or an agent misbehaves, it can only reach what it was granted. For AI agents it is the main lever against power seeking and prompt-injection damage — narrow the permissions and you narrow the harm. It is foundational to defense in depth.
For product teams
The cheapest way to shrink how bad any single failure can get.
For engineers
Scope every credential and capability to the minimum; it bounds blast radius from compromise, injection, or a misaligned agent.
Related
- Defense in depth — A core layer of defense in depth.
- Power seeking — A key check on power seeking.
- Prompt injection — Limits damage from prompt injection.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome