AI Trust and Safety User Guide

Welcome to the AI Trust and Safety User Guide, an introduction to the broad topic of trustworthy, safe AI, prepared by The AI Alliance.

Tips:

1. Use the search box at the top of this page to find specific content.
2. Capitalized Terms link to glossary definitions.

This guide is organized as follows:

• Introduction to Trust and Safety: What these concepts mean to us and why they are important.
• Glossary: How we define various terms.
• Exploring AI Trust and Safety: Several expert explorations of key concepts. This is the heart of the guide:
  • NIST Artificial Intelligence Risk Management Framework: A framework developed by the National Institute of Standards and Technology, under the United States Department of Commerce.
  • Trust and Safety at Meta: R&D at Meta on trust and safety.
  • Mozilla Foundation’s guidance on Trustworthy AI: Mozilla Foundation’s guidance on ensuring trustworthy AI.
  • MLCommons AILuminate: The influential risk taxonomy and corresponding benchmarks from the MLCommons industry collaboration.
  • The Trusted AI (TAI) Frameworks Project: Trustworthiness research from an academic and armed forces collaboration.
  • Cybersecurity: New security considerations when using AI.
• Safety for Your AI Systems: Some particular recommendations on how to successfully build trusted and safe AI systems.
• Final Thoughts.
• References: For more information.

Help Wanted! We want to expand the content in Exploring AI Trust and Safety, and greatly expand the information throughout on how you can apply what you learn. We need your help!

Additional links:

• Contributing to the User Guide: We welcome your contributions! Here’s how you can help.
• About Us: More about the AI Alliance and this document.
• The AI Alliance
• This Guide’s GitHub Repo

Version History

---