Skip to content
⌘K
InboxAsk a human

Introduction to Unbox Safety

Updated by Hugo

Unbox Safety monitors shoppable content produced by creators for your products, and enables you to action when something goes wrong, using advanced multimodal AI and automated enforcement.

This article explains what the safety system does and how its parts fit together.

How Safety works

When a creator posts a video tagged with a shoppable link to your product(s), Unbox AI analyses the content against the risks you've configured. If a risk is detected, the system either acts immediately or flags the video for manual review, depending on how you've set it up.

For videos running in GMV Max campaigns, you can enable Quarantine. This is a setting rather than a policy: when it is on, any video with an unresolved flag is held out of paid amplification from the moment the risk is detected, until the risk is cleared. This protects your spend while the review is under way.

Where Safety lives

Safety sits in three places in your navigation: two under Videos, one under Insights.

  • Flagged Videos – the review queue, where flagged videos are accepted, dismissed and tracked. Under Videos, alongside All Videos.

  • Configure Safety – your risks, severities and policies, plus the Quarantine settings. Also under Videos.

  • Safety Overview – reporting: what needs attention, your risk profile, and what enforcement prevented. Under Insights.

The parts of the system

Risks

A risk is a specific type of content problem Unbox can detect, for example an unsubstantiated product claim, use of unlicensed music or explicit content. Unbox provides a library of standard risks grouped by category. You choose which ones are active for your brand and configure how each one behaves.

Our core safety risks are always-on and cannot be disabled. These cover content that no brand should be amplifying regardless of context: violence, sexual content involving minors, hate, and similar.

→ Risk catalogue

Risk configuration

Risk configuration is where you set up how Unbox responds to each risk. For every active risk you decide: how serious is it (severity), which policy should run when it's detected, and whether that policy should run automatically or wait for a reviewer to confirm the flag first.

Admins can edit the risk configuration using the Configure Risks button on the Risk configuration page. Standard users can view the configuration, but aren't able to make changes.

→ Setting up your risk configuration

Policies

A policy is what happens when a risk is confirmed: a set of steps that runs on its own until the risk is resolved. Policies range from sending the creator a friendly warning all the way to ending the relationship entirely. When multiple risks fire on the same video, the strictest applicable policy runs. Only one policy executes per video.

→ How risk policies work

The review queue

The review queue is where your team handles flagged videos that need a human decision. It has three tabs:

  • Needs Review – flags awaiting acceptance or dismissal, populated only for risks set to Require review.

  • With Creator to Action – videos where a policy is running and waiting on the creator to respond.

  • Risk Resolved – a complete record of all closed risks.

→ How to review & action risks

Reporting

The Safety Overview brings together what needs your attention now, the shape of your risk over the last 30 days, and what enforcement has prevented.

How a flagged video moves through the system

  1. Unbox AI detects a risk on a creator's video.

  2. If the risk is set to Auto-approve, it is confirmed immediately and the policy assigned to that risk runs. If set to Require review, it appears in the Needs Review queue.

  3. A reviewer accepts or dismisses the flag. The policy assigned to that risk is already selected and shown under “What happens next”, so you can see what will happen before you save. You can choose a different policy if you disagree with the suggested one.

  4. Accepting the risk runs that policy. Its steps then happen automatically: contacting the creator, removing the video from GMV Max campaigns, or ending the collaboration.

  5. Once the policy completes, the risk is closed and moves to Risk Resolved.