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

# Data Privacy

> How Velatir handles session content by default, what stays after review, and how you choose storage and retention.

## Privacy by Default

Velatir starts from **Never**: the raw prompt and response are not stored. You still get a generalized description of the interaction, the agent outcome, and a full audit trail. An organisation admin can keep raw content when it is needed — always, or only when an agent flags the interaction — and choose how long records live.

## What Happens to an Interaction

<Steps>
  <Step title="It is captured" icon="activity">
    When someone uses an AI service Velatir monitors, the interaction is sent to Velatir so your agents can review it. Velatir does not collect browsing history, cookies, credentials, or local files.
  </Step>

  <Step title="Agents review it in memory" icon="shield-check">
    Your agents assess the prompt in flight — in memory — and allow, flag, or block it. The raw wording is not written to storage for that review.
  </Step>

  <Step title="What is kept depends on you" icon="database">
    A generalized description is always stored. Raw prompts and responses are written to storage only if you choose **When flagged** or **Always**. See [Data storage & encryption](/security/data-storage-and-encryption) for how that stored data is protected.
  </Step>
</Steps>

<Note>
  Under **Never**, the default, the raw prompt never hits disk. It is processed in memory and discarded after review. **When flagged** and **Always** are the opt-ins that persist it.
</Note>

## Session Content

**Store trace data in sessions** controls when raw prompts and responses are kept. The default is **Never**.

| Setting             | What is stored                                                                                                  |
| ------------------- | --------------------------------------------------------------------------------------------------------------- |
| **Never** (default) | Raw content is never written to storage. Review happens in memory. The Context tab shows no prompt or response. |
| **When flagged**    | Raw content is stored only for interactions an agent flags. Everything else is treated like **Never**.          |
| **Always**          | Raw content is stored for every interaction.                                                                    |

## Generalized Summaries

Every interaction gets a short, generalized **Description**, even when raw content is not stored. That is what you see in the session list and on the Summary tab.

The description says what *kind* of request it was, not the actual words. Names, companies, numbers, identifiers, URLs, and verbatim excerpts are replaced with generic terms — for example, “The user is asking the assistant to draft an email to a colleague.”

<AccordionGroup>
  <Accordion title="What the summary is for" icon="align-left">
    It keeps the session list useful under **Never**, and it is the record of the interaction when you choose not to keep the raw prompt.
  </Accordion>

  <Accordion title="What it does not include" icon="eraser">
    It is written to drop specifics. It is not a copy of the prompt, and it is not covered by a customer-held encryption key — those keys protect stored Context, not the Description.
  </Accordion>
</AccordionGroup>

## How Long Data Is Kept

**Data retention period** controls how long session records live — the description, assessments, and any raw content you chose to keep. The default is **Always**.

| Setting                                                      | What happens                                              |
| ------------------------------------------------------------ | --------------------------------------------------------- |
| **Always** (default)                                         | Records are kept until you delete them.                   |
| **6 months**, **12 months**, **24 months**, or **36 months** | Records older than that window are deleted automatically. |

Changing from **Always** to a bounded period starts a seven-day grace period, so a misclick does not wipe history immediately. Settings shows the next deletion date; change the period before that date to prevent it.

Storage mode and retention are independent: one decides *whether* raw content exists, the other decides *how long the record lives*.

## Configure It

<Steps>
  <Step title="Open organisation settings" icon="settings">
    Go to **Settings → General** and find the **Data & privacy** section.
  </Step>

  <Step title="Choose when to store raw content" icon="eye">
    Set **Store trace data in sessions** to **Never**, **When flagged**, or **Always**.
  </Step>

  <Step title="Choose how long to keep records" icon="calendar">
    Set **Data retention period** to **Always**, or to 6, 12, 24, or 36 months.
  </Step>
</Steps>

## What Is Recorded

Velatir keeps an internal record of account activity so there is a trail for reviews and support. That includes member invitations and role changes, organisation and workspace settings updates, agent configuration and instruction changes, and who was notified of a high-criticality finding.

This record is not a dashboard you can browse today. What you work from day to day is **Sessions**, **Assessments**, and [data exports](/security/data-storage-and-encryption#data-exports). Channel notifications do not have an approve/decline workflow — see [Escalations](/traces/escalations).

## Support Access

Velatir's team has no access to your account by default. No one at Velatir can view your data, traces, or settings unless you grant access under **Settings → Support**, and revoking it takes effect immediately.

## EU-Based Infrastructure

Velatir runs entirely on EU-based infrastructure. Your data stays within the European Union, which supports data residency and sovereignty requirements and aligns with GDPR. There are no data transfers to non-EU jurisdictions.

***

<CardGroup cols={2}>
  <Card title="Data storage & encryption" icon="lock" href="/security/data-storage-and-encryption">
    What is stored, and how it is protected.
  </Card>

  <Card title="Roles & permissions" icon="shield" href="/platform/roles-and-permissions">
    Control who can see and change what.
  </Card>

  <Card title="Redaction" icon="eraser" href="/agents/redaction">
    Strip sensitive content from pasted text, on the device, before it is sent.
  </Card>
</CardGroup>