---
title: "Privacy and consent"
description: "Keep personal information out of frame, know what to do if an adult enters, and never record a minor."
---

<span className="otz-eyebrow">Recorder playbook · Keep every clip clean</span>

The easiest rule is simple: keep other people and personal information out of frame when
you can. If an adult unexpectedly enters a recording in Texas, there is a straightforward
way to ask for permission without restarting the whole task.

## The three rules to remember

<CardGroup cols={3}>
  <Card title="No minors" icon="child-reaching">
    Never record a child. If someone may be under 18, stop and keep them out of the clip.
  </Card>
  <Card title="Keep PII out" icon="id-card">
    Avoid badges, plates, addresses, screens, documents, mail, and other identifying
    details.
  </Card>
  <Card title="Ask adults" icon="comment-check">
    An adult who appears needs a signed release or, for a truly incidental Texas
    bystander, a recorded verbal yes.
  </Card>
</CardGroup>

## If an adult walks into frame in Texas

This quick option is only for an **incidental adult bystander**, meaning someone who is
not the recorder, not part of the task, and not being featured or followed by the camera.

<Steps>
  <Step title="Ask right away" icon="message">
    Keep the camera recording long enough to capture the question and answer. Say:

    > “This camera records video and audio for opentez. opentez may use the footage to
    > develop AI systems and may include it in commercial AI-training datasets provided
    > to other companies. Are you okay with being recorded and with opentez using the
    > footage that way?”
  </Step>
  <Step title="Listen for a clear yes" icon="volume-high">
    The adult must give an audible, unambiguous **yes**. A nod, silence, or simply staying
    in the room is not enough.
  </Step>
  <Step title="Mark the consent" icon="clock">
    Note the clip name and the exact timestamp of the question and answer. Tell your SPL
    that you used the Texas verbal-bystander script.
  </Step>
  <Step title="Use only what follows" icon="scissors">
    Footage from before the disclosure and answer does not qualify. Continue only after
    the recorded yes.
  </Step>
</Steps>

<Tip>
  If the person says no, seems unsure, or may be under 18, stop recording and continue
  after they leave. That is all you need to do.
</Tip>

## When you need the signed Subject Release

Use the full **Subject Release** when the person:

- is deliberately taking part in the task;
- is featured, interviewed, followed, or repeatedly visible;
- will be singled out in a sample, description, or individual annotation;
- is being recorded outside Texas, or the recording state is uncertain; or
- is covered by a project that specifically requires a signed release.

The verbal option never covers facial recognition, voiceprints, identity matching, or
other biometric identification. It also does not permit recording private conversations.
Keep private conversations out of every clip.

## Before you upload

Submit every required consent record with the session. For a Texas verbal bystander, that
means the clip containing the full question and audible yes, plus the clip name, consent
timestamp, and script version given by your SPL.

<Note>
  When you are unsure which option applies, use the signed Subject Release or ask your
  SPL. Both are quick, and either one keeps the footage easy to accept.
</Note>

## Why this matters

opentez accepts footage with a clear, documented chain of consent. A simple record of who
agreed, what they agreed to, and where that consent appears keeps the footage useful and
easy to verify.

<Card title="What makes a clip valuable" icon="gem" href="/what-were-building#what-makes-a-clip-valuable">
  The full picture of why clean provenance matters most.
</Card>