Checklist

AI Safety Pre-Check Checklist for Prototypes

An operational interactive pre-check teams can complete before launching AI prototypes. Collects yes/no responses, evidence notes, risk flags, ownership, and suggested next steps so teams can act, escalate, and track follow-up.

Interactive Tool

AI Safety Pre-Check Checklist for Prototypes

Use this short pre-check before exposing an AI prototype to users. The goal is to identify obvious safety, privacy, bias, or compliance problems, capture evidence, and record owners and immediate actions. Be honest — this checklist helps teams catch and fix issues early, not block safe experiments. Save the pre-check and escalate any 'No' or high-risk answers per your governance process.

Can you trace model inputs to source datasets, extraction dates, transformations, and owners? Provide links or IDs in the notes.
Dataset names, ETL scripts, storage locations, dataset versions, or documentation links.
Has personal data been identified and treated according to policy?
Describe what personal data exists, how it is minimized, access controls, and retention plans.
Have you verified that no examples or labels leaked between training and evaluation sets and that features do not trivially reveal targets?
Does the class distribution risk biased outputs for particular groups?
Describe plausible failure scenarios, who is impacted, and severity (e.g., wrong diagnosis, biased denial, privacy leak).
Is there a brief, user-friendly explanation of the prototype’s purpose, limitations, and data/consent use?
Specify whether and when a human needs to review or approve outputs (e.g., all high-risk cases).
Is there a tested plan (and person/permission) to pause or roll back the prototype if problems appear?
List metrics, alerts, thresholds, and where alerts will be sent (e.g., model confidence, distribution drift, user complaints).
Select applicable issues that have been escalated to legal or compliance.
Rate readiness to launch this prototype to users based on checklist results.
1.0 10.0
Name and role of the person approving this pre-check.
Capture actions, owners, and target dates for any 'No' or high-risk items.
If any critical issues exist, select Yes to trigger escalation.
You can explore this tool now. Sign in or create an account to save your responses and return to them later.
Sign In to Save Create an Account
Make this tool part of your work

Save a personal copy, bring it to your team, or tailor the questions and workflow to fit what you are hungry to improve.

Member customization and team collaboration are coming soon.
Discuss Ask the Librarian

Discussion

Comments and conversation will live here.