> ## Documentation Index
> Fetch the complete documentation index at: https://docs.exterview.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Human Oversight

> The control model Smaya is designed around from its first release: read-only by default, with any future ability to act gated behind explicit approval.

## Purpose

Smaya hasn't shipped yet. Its first release won't be able to take an action or change data, so there's nothing to gate yet, but the control model is designed in from the start, before any action capability ships.

## The control model

<CardGroup cols={2}>
  <Card title="Read-only in the first release" icon="lock">
    Smaya's first release will answer questions with citations. It won't be able to start a session, change a record, or take any action.
  </Card>

  <Card title="Approval before any future action" icon="user-check">
    If Smaya ever gains the ability to act on your behalf, that action is designed
    to require your explicit approval first, nothing changes automatically.
  </Card>

  <Card title="Specific to your organization" icon="building-shield">
    Every answer is grounded only in your organization's own knowledge and the
    record on screen, never another customer's data.
  </Card>

  <Card title="Redacted by role, automatically" icon="eye-slash">
    Compensation figures and internal-only signals — integrity flags, private
    panel notes — are filtered out of a hiring-manager or candidate answer before
    it's shown, not something anyone configures per question.
  </Card>

  <Card title="Logged for review" icon="clipboard-list">
    Every question, answer, and citation is recorded, so your compliance team can review what was asked and where the answer came from.
  </Card>
</CardGroup>

<Warning>
  No autonomous adverse decisions, now or planned. Smaya is not designed to ever
  reject or disadvantage a candidate on its own, that judgment stays with a
  person, the same standard held everywhere else in Exterview. See [Responsible
  AI](/trust/responsible-ai).
</Warning>

## Why it's built this way

Designing the approval gate before any action capability exists, rather than bolting it on afterward, is what makes it trustworthy for regulated, enterprise hiring: the control model isn't a promise to add later, it's the starting point.

## FAQs

<AccordionGroup>
  <Accordion title="Can Smaya change our data today?">
    No. Smaya hasn't shipped yet, and its first release will be read-only — it will answer questions and won't be able to take actions.
  </Accordion>

  <Accordion title="If Smaya can act in the future, who approves it?">
    You do. The design requires explicit approval before any action executes,
    nothing is designed to happen automatically.
  </Accordion>

  <Accordion title="Is there a record of what Smaya was asked?">
    Yes. Every question and answer is logged for your compliance team to review.
  </Accordion>
</AccordionGroup>

## Related

<CardGroup cols={2}>
  <Card title="Meet Smaya" icon="compass" href="/smaya/overview">
    What Smaya is, and what it isn't yet.
  </Card>

  <Card title="Responsible AI" icon="scale-balanced" href="/trust/responsible-ai">
    The same human-review standard applied platform-wide.
  </Card>
</CardGroup>