Continuous AI assurancePrivate preview available by request

    An approval is only as current as the system it approved.

    Models are upgraded, prompts are edited, tools are added. NVX is designed to show which approvals and evidence are affected by a change, so reviews happen when they are needed rather than once a year.

    What goes stale

    Changes that can quietly invalidate an approval.

    Model version

    A new model can behave differently on the same inputs.

    Prompts and instructions

    Small wording changes can alter outputs and safety behaviour.

    Tools and permissions

    New tools extend what an agent can do.

    Data sources

    New or changed data affects accuracy, privacy and fairness.

    How NVX approaches it

    Versioned records that show what changed.

    Versioned assurance records are planned for the private MVP. The goal is that each approval references the exact configuration it covered, and a later change marks the affected evidence for review.

    1. 01

      Record the version

      Capture model, prompt, tool and data configuration at approval.

    2. 02

      Detect the change

      Register a new version when any of those change.

    3. 03

      Flag affected evidence

      Show which controls and evidence relied on the previous version.

    4. 04

      Re-approve

      Route the system back to its owner for a recorded decision.

    Questions

    Frequently asked

    Is continuous assurance available today?
    Not fully. Versioned assurance records are planned for the private MVP and change detection is under design with partners.
    Does every change need a full review?
    No. Your organisation decides which changes are material. The aim is to make that decision visible and recorded.
    How does this relate to model monitoring?
    Monitoring tools watch live behaviour. Continuous assurance connects what monitoring finds to the risks, controls and approvals that depend on it.

    See the workflow on one of your AI systems.

    NVX is accepting selected enterprise teams into a private preview of its AI governance and assurance workflow.