Agent identity
Also called AI agent identity, identity for an AI agent
An account made for one AI agent, separate from any person's login, so its actions can be traced and it can be switched off alone.
Every action an AI agent takes runs on somebody's permissions. An agent identity makes that somebody the agent. It is an account created for one agent, with its own credentials, holding only the access the agent's job needs.
Without one, an agent borrows. It runs on the personal token of whoever set it up, or on a shared account other automation already uses. Then the log blames a person for what the agent did, the agent can do everything that person can, and stopping it means locking out someone else.
With one, the log names the agent, its access can be cut to the task, and it can be disabled alone. A register of agents starts from these identities: one line each, with an owner.
For example
The log reads "triage-agent closed 40 tickets", not the name of the engineer who built it, and disabling triage-agent stops that one agent and nobody else.
Put it to use
Where you'll meet agent identity in ChatGPT, Gemini and Claude, how to use it there, and the mix-up most people make.
Daily Bytes Plus is in the Android app.