Agent identity and authorisation

Definition

The mechanisms used to establish which AI agent is acting and what that agent is permitted to access or do.

Put simply

If an agent is working inside your organisation, it increasingly needs something equivalent to its own identity and permissions.

What boards and leaders should understand

As agents gain access to email, customer information, financial systems, files and other applications, identity and access management becomes part of AI governance.

A particularly important distinction is whether an agent acts through a person’s existing credentials or has its own identifiable authority.

Separate agent identities can improve attribution, limit permissions and make it easier to establish which agent performed an action.

Questions to consider

  • Can we identify which agent performed an action, on whose behalf it was acting and what authority it had at the time?