AI Capabilities That Can Take Actions

These Registry records describe AI capabilities for which current evidence establishes the ability to change data, configuration, workflow state, or another operational outcome. This view is generated from the Registry’s current capability records. It does not rank products or assess whether an action is beneficial, harmful, safe, or appropriate.

“Can take actions” does not mean that a capability operates without human involvement. Some capabilities require confirmation, some operate conditionally, and for others the confirmation requirement is not publicly established.

How the Registry classifies capability behaviourView capabilities by human confirmationView capabilities by permission basis

A capability appearing in this view means the Registry has evidence supporting its action-taking classification. It does not establish that the capability acts autonomously, that it acts without confirmation, that it can act outside its own product, that it has unrestricted permissions, or that it is enabled by default. Those properties are recorded separately where evidence permits.

84 capabilities currently recorded