AI risk and controlsPrivate preview available by request

    Every AI risk should point to a control, and every control to evidence.

    NVX records risks against the specific AI system they apply to, maps each risk to the controls intended to address it and keeps the evidence that those controls work.

    Common AI risk categories

    Risks that recur across enterprise AI systems.

    These categories are a starting point. Your organisation decides which apply to each system and how they are rated.

    Privacy and data leakage

    Personal or confidential information exposed in prompts, outputs or logs.

    Inaccurate or harmful output

    Answers that are wrong, misleading or inappropriate for the audience.

    Unauthorised action

    An agent taking an action outside its approved scope.

    Unfair outcomes

    Different treatment of people or groups that the organisation cannot justify.

    Dependency failure

    A model, vendor or data source changing or becoming unavailable.

    Lack of transparency

    People affected by a decision cannot tell that AI was involved or how.

    How it is recorded

    From risk to control to evidence.

    1. 01

      Describe the risk

      In terms specific to the system, with a likelihood and impact your framework defines.

    2. 02

      Map the control

      Link one or more controls and record who operates them.

    3. 03

      Attach evidence

      Test results, configurations or review records, each dated and sourced.

    4. 04

      Record findings

      Where evidence is missing or a control fails, log a finding with an owner and due date.

    Questions

    Frequently asked

    Do I have to use NVX's risk categories?
    No. The categories above are examples. You can record risks using your organisation's own taxonomy and rating scale.
    Can one control cover several AI systems?
    Yes. A shared control can be linked to many systems, but evidence that it operates is kept per system, because the same control can work in one place and fail in another.
    Does NVX rate risk for me?
    No. NVX records the ratings your reviewers assign and the reasoning behind them.

    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.