AI Safety Review

Review flagged AI outputs against safety policies with category, severity, source context, model trace, reviewer notes, and final disposition. This example brings Flagged and High severity together so AI product teams can find the next task without losing context.

It includes representative records such as Unsafe instruction escalation and Sensitive-data exposure risk. Use the screenshot to study the layout, then adapt its components and product states to your own design system.

AI Safety Review
Use this template in Figr

Template facts

Screen

Inbox

Primary action

Resolve review

Pattern

Inbox

State

Default, loading, empty, error, permission, success

Design system

Configurable product UI theme

What this screen helps users do

Product teams need to resolve review while keeping flagged, high severity and exceptions visible. The interface should show what changed and who should act next.

Key elements

  • Main workspace: Flagged gives users an obvious place to start.
  • Supporting view: High severity keeps the related context close.
  • Primary action: Resolve review stays visible when people need it.
  • First signal: Flagged shows a useful value (18).
  • Second signal: High severity gives additional context (3).
  • Sample record: Unsafe instruction escalation.
  • Next record: Sensitive-data exposure risk.

States worth designing

  • No items match the selected filter: say what changed, protect progress, and offer a clear next step.
  • Message or task loading: say what changed, protect progress, and offer a clear next step.
  • Reply or assignment failed: say what changed, protect progress, and offer a clear next step.
  • Item resolved: say what changed, protect progress, and offer a clear next step.
  • Escalation required: say what changed, protect progress, and offer a clear next step.
  • Permission to read restricted content missing: say what changed, protect progress, and offer a clear next step.

Where this screen sits in the flow

Flagged → High severity → Resolve review → Review outcome

Patterns used

Inbox, Flagged, High severity, contextual actions, error recovery

Why these choices work

Keep the current item close to its history, owner and urgency. Put the next action on the record rather than making people open several panels.

Pay attention to flagged and high severity. People need to distinguish Unsafe instruction escalation from Sensitive-data exposure risk. Keep flagged visible, especially when a record changes owner or status.

Before this screen reaches handoff, verify keyboard focus, empty and permission states, and what happens after someone chooses “Resolve review”. Make error recovery as clear as the happy path.

Related screen examples to explore next: Human Review Queue, AI Output History, Evaluation Dataset.

Start from this in Figr

Design a ai safety review screen inside our existing product. Use our current components, tokens and content language. Include Flagged and High severity views, credible records like Unsafe instruction escalation, and a clear “Resolve review” action. Use our product permissions and show loading, empty, error and success states. Keep the output editable in Figma.

Open Figr