AI system review

Make the purpose, risks, evidence, and decision inspectable.

Start with a bounded use, gather evidence for the questions that matter, and give reviewers a case they can inspect. Continue with recorded conditions after the decision.

Read the guide

Frame the review

Define the intended use and its boundary. If a standard is relevant, select its edition and examine which requirements apply.

Review an AI system before a pilot

Build a case for a bounded use before deciding whether to proceed.

Assess risks

Find gaps against a selected standard

Turn relevant clauses into evidence work with a visible source.

Assess risks

Examine the evidence

Establish a clear Dataset basis, then investigate a supported condition that matters to the intended use. Keep the settings and limits with the finding.

Prepare a Dataset for testing

Bring samples and labels into Vault with a clear task and version.

Prepare data

Test a model under changing conditions

Compare the clean reference with conditions that matter to the intended use.

Test models

Hand off and follow up

Let the next reviewer open the same saved case. After approval, record observations, condition status, and the reasons to revisit the decision.

Hand an Assessment to another reviewer

Share the saved case, exact evidence, and the work that remains.

Share evidence

Follow up after an approval

Keep observations and conditions connected to the approved use.

Share evidence