Skip to main content

Trust Auditor

Pick one AI feature (a summariser, an agent, an autocomplete) and answer each question about it honestly. Answers save automatically in your browser; nothing leaves your device.

  1. 01

    Visible confidence

    Signal how sure the system is, and mark the parts it is least sure about.

    Does the interface communicate the system's confidence, or flag output it is uncertain about?

  2. 02

    Reviewable output

    Let people see and inspect what the AI produced before it is used or sent.

    Can the user inspect the AI's output before it takes effect?

  3. 03

    Cheap correction

    Allow editing or overriding output in place; never force a full restart to fix one thing.

    Can the user edit or override the output without starting over?

  4. 04

    Graceful recovery

    Provide a clear way back when the AI is wrong: undo, retry, or escalate.

    Is there a clear recovery path when the output is wrong?

  5. 05

    Human control on consequential actions

    Require explicit confirmation before anything irreversible or high-impact.

    Do consequential or irreversible actions require explicit human confirmation?

  6. 06

    Honest expectations

    Don't imply certainty or ability the system lacks; label AI-generated content.

    Is AI-generated content labelled, and does the framing avoid implying false certainty?

  7. 07

    Preserved context

    Don't silently drop what the user provided; keep their inputs visible and honoured.

    Does the system keep and honour the context the user provided, visibly?

  8. 08

    Shown provenance

    Make output verifiable: show sources, inputs used, or how it was produced.

    Can the user see where the output came from, or what it was based on?