Ethical AI

When an AI system causes harm, the law looks for a person. Not a model.

In 2024 a tribunal held Air Canada responsible for what its chatbot told a customer. "The algorithm did it" did not survive contact with the bench. That precedent is now one of dozens, and the rules are arriving fast: the EU AI Act, ISO 42001, a growing patchwork of national laws.

Ethical AI is a working guide to governing AI before it governs you. It is written for the person who owns the P&L, signs the procurement, or carries the legal exposure when something goes wrong - usually not technical, usually short of time, increasingly on the hook.

Inside:

• Who actually answers for an AI system, and how to assign it on paper
• The Three Lines of Defence, applied to AI
• An AI risk register and impact assessment you can run
• The EU AI Act and the US patchwork, in plain terms
• What to do when a model goes wrong: incident response that holds up
• A 90-day governance starter

This is not a lecture on AI ethics. It is the practical work of not blowing up in your face, and the governance that turns out to be a competitive advantage.

By Sotiris Spyrou. Part of the Executive AI Mastery series, alongside TRANSFORM and AI Moats.