Ethics Guidelines for Trustworthy AI

Why it matters

Where the EU AI Act tells you what is legally required, this guide is the earlier document that tells you what "trustworthy" was meant to mean in the first place - and its assessment checklist remains one of the most usable self-audit tools available for free.

Key takeaways

• Defines trustworthy AI as lawful, ethical and robust - three distinct, jointly-necessary conditions, not a single score.
• Translates four ethical imperatives (respect for human autonomy, prevention of harm, fairness, explicability) into seven concrete requirements, including human agency and oversight, technical robustness and safety, and accountability.
• Includes a practical Assessment List for Trustworthy AI (ALTAI) with more than 60 self-assessment questions organisations can run against a specific AI system.
• Not legally binding - it predates the EU AI Act by five years and was explicitly piloted with industry before any legislation was drafted.

Part of these reading paths

• Understanding AI Governance
• AI for Business Leaders