Stockertim Dictionary · Artificial intelligence · Cybersecurity
Non-human identity
Also called machine identity, workload identity, service account
In short
A non-human identity is an account, key or credential used by software, such as an AI agent or automated script, rather than by a person.
- Advanced
- Concept
- Checked
In detail
Non-human identities are the log-ins of software. When an AI agent reads a mailbox, a script updates a database or two systems exchange data, they authenticate with an account, a key or a token, just as a person signs in with a password. These identities often outnumber human accounts and can hold broad, long-lived access.
They need care because they lack many of the protections people have. A software key does not use a second factor in the same way, does not leave the company, and may never be reviewed. Keys can end up in shared documents or code, accounts can keep permissions long after a project ends, and one identity shared across several tools makes it impossible to tell which one did what.
Good practice applies least privilege and clear ownership. Each AI agent or tool gets its own identity with only the permissions its task needs, a named human owner, an expiry or rotation date for its keys, and logging of everything it does. Sensitive actions still require a human approval step. When a tool is retired, its identity is switched off and its keys revoked.
Regular access reviews include non-human identities alongside staff accounts, checking that each one is still needed, still owned and still limited to what it uses.
Worked example
Kite & Co Accountants runs an access review of the identities used by its AI tools.
- Non-human identities found: 12
- Not used in the last 90 days: 5, so 5 / 12 = 41.7%
- Shared by more than one tool: 2, which are split into separate identities
- After switching off the 5 unused ones and splitting the 2 shared ones: 12 − 5 + 2 = 9 identities
The firm ends with 9 identities, each used by one tool, owned by a person and limited to its task.
In practice
IT and security teams give each AI agent its own identity, store keys in a secrets manager, rotate them on a schedule, log activity and include machine accounts in quarterly access reviews.
Watch out
- Giving an agent a staff member's own log-in hides what the agent did and gives it all of that person's access.
- A key that never expires is a standing risk even if it has never been misused.
Learn it properly
Official sources
- Guidelines for secure AI system development · National Cyber Security Centre
- 10 Steps to Cyber Security · National Cyber Security Centre
- Glossary · NIST Computer Security Resource Center
Facts last checked on 3 November 2026.
Education, not advice: the Stockertim Dictionary explains what terms mean. It never recommends buying or selling anything.