Humanitarian AI Risk Screen

About this toolkit

This resource helps you decide whether and how to use an AI tool for a humanitarian task. It includes:

• A seven-step risk screen with accountable reviewers
• A pilot charter template for amber outcomes
• Synthetic training case cards
• A source-check ledger and curriculum source matrix
• A command-line tool (ai-risk-screen) for validation and reporting

Disclaimer: This is an open educational resource informed by cited humanitarian and privacy instruments. It is not an official standard, legal advice, or your organisation's policy. All case scenarios are synthetic.

Outcomes

Green

Approved low-risk assistance with mandatory source checking

Amber

Pause for privacy, protection, or partner review; proceed via controlled pilot only

Red

Do not use — prohibited input, unapproved tool, unreviewable high-stakes decision, or no safe remedy

Quick start

1. Read the seven-step risk screen
2. Complete the browser scorer (no data leaves your device)
3. Or install the CLI: pip install humanitarian-ai-risk-screen