# Understanding alerts (/en/docs/platform/dashboard/alerts)



An alert is created when the child app detects a risk signal above its threshold. Only `medium` and `high` risk levels create an alert — weaker signals are discarded on the child's device. SafeCircle never sends the message text.

## Risk levels [#risk-levels]

<Tabs items="[&#x22;High&#x22;, &#x22;Medium&#x22;]">
  <Tab value="High">
    A strong risk signal. Shown with a **red** badge. Worth a conversation the same day.
  </Tab>

  <Tab value="Medium">
    A moderate risk signal. Shown with an **orange** badge. Worth a look — it may be a false positive depending on context.
  </Tab>
</Tabs>

## What an alert shows [#what-an-alert-shows]

Open an alert from the list to see its details.

| Field            | Description                                                                                                                                                          |
| ---------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Risk level**   | High or medium, plus a **Read** / **Unread** badge                                                                                                                   |
| **Child**        | Which child's device it came from                                                                                                                                    |
| **Source app**   | The app where the signal was detected                                                                                                                                |
| **Confidence**   | How sure the model is, from 0 to 100%                                                                                                                                |
| **Detected**     | When it was detected                                                                                                                                                 |
| **Categories**   | One or more risk labels from the model                                                                                                                               |
| **AI Reasoning** | A short note on why it was flagged. For now this is a plain summary of the detected categories (for example, *Detected: grooming*). It never quotes the conversation |

## Finding alerts [#finding-alerts]

The **Alerts** page has four tabs:

* **All** — every alert, newest first.
* **Unread** — only alerts you haven't opened yet, with a count.
* **High** and **Medium** — filter by risk level.

If you have more than one child, a **Filter by child** menu appears. Pick **All children** to see everyone again.

A summary panel shows **Total alerts**, **Unread**, **High risk**, **Medium risk** and **Top app** for the current selection.

You can also open an alert directly from the bell in the top bar, which lists the latest five alerts.

## Marking alerts as read [#marking-alerts-as-read]

Opening an alert marks it as read. You can also:

* Click **Mark as read** in the alert's details.
* Click **Mark all read** above the list to clear everything unread in the current view.
* Use **Mark all read** in the bell menu.

Read alerts stay in the **All** tab and in the [Timeline](/en/docs/platform/dashboard/analytics-timeline#timeline). Nothing is deleted.

## What to do next [#what-to-do-next]

The Alerts page includes a short guide, **What to do next**:

* Start with curiosity, not punishment. Ask what happened before talking about the app.
* High risk alerts are worth a conversation the same day.
* Confidence is a rough indicator, not a probability. Even a lower score can be worth a look.

## Related [#related]

<Cards>
  <Card title="Risk categories" href="/en/docs/ai/risk-categories" description="What each category, risk level and the reasoning note mean." />

  <Card title="Notifications" href="/en/docs/platform/dashboard/notifications" description="Get alerts as they arrive, in the dashboard or the parent app." />

  <Card title="Analytics and Timeline" href="/en/docs/platform/dashboard/analytics-timeline" description="See patterns across days, categories and apps." />

  <Card title="FAQ and troubleshooting" href="/en/docs/support/faq" description="Not receiving alerts? Work through the checklist." />
</Cards>