Models

    The LLM an AI agent answers through determines where prompts and enterprise data are processed, which provider terms apply, and what the agent can do. Each agent is configured with its own model, visible only inside the platform that hosts it. Models in SecureAI collects the models from every AI platform you have onboarded into. Use Models to establish which models are in service, which providers your agents depend on, and which of those models carry live agent workloads. The inventory only reports those models; it is not where a model is changed.

    How it works

    Every model in the inventory is added via a sync rather than created in SecureAI. A sync runs against an onboarded instance and records the models that instance offers to its agents, along with the agents already configured for each model. The inventory is as current as that platform's instance's completed sync, which runs every six hours by default.

    A model appears because an onboarded instance reported it, and the agents shown against it are only those the same sync discovered. An instance that has not completed a sync contributes neither its models nor its agent usage.

    Prerequisites

    The following must be in place before the model inventory returns records.

    • Your SecureAI license must cover Governance.
    • To view the inventory, log in as a Super Admin or as a help desk technician holding a role that includes the Models task under AI Inventory. The default Governance Manager role includes it.
    • Add at least one AI platform under Application Settings and associate the service that reports its agents.
    • Sync each instance and confirm the sync has completed.
    Note:
    • The inventory offers no export capabilities.
    • A change reaches the inventory only after the next sync for that instance is completed. A model newly enabled for an instance and an agent newly switched to a different model will both reflect the state they held at the last sync.
    • Models belonging to an AI platform that is not yet onboarded are not discovered.
    • A model that no discovered agent is using is still listed, so the inventory is not a record of models in active service.

    What the inventory reports

    Models reports the following against each discovered model:

    Column What it shows
    Model The name of the discovered model.
    Model ID The identifier held against that model.
    Provider The provider that supplies the model, such as OpenAI or Anthropic.
    Application The onboarded application that reported the model, such as Copilot Studio, Microsoft Foundry, or Zia.
    Instance The application instance that reported the model, where more than one instance of the same application is onboarded.
    Used By The agents configured to answer through the model.