Ethical AI

In February 2024 a Canadian tribunal held Air Canada liable for what its chatbot had told a passenger. The airline argued the chatbot was a separate legal entity. The tribunal disagreed. Within eighteen months the same question was being asked in courts in the European Union, the United States, Singapore, and Australia.

Ethical AI is the practical governance reference for the executive who has to answer when the model goes wrong. Not a philosophy book. Not an academic survey. A working guide for the board director, C-suite executive, general counsel, or risk officer whose organisation has deployed AI faster than it has built the apparatus to govern it.

Eighteen chapters across five parts:

• The governance problem - accountability, the regulatory map, what "ethical AI" actually means
• How AI systems fail - bias, explainability, privacy, agentic and safety-critical systems
• Building a governance posture - risk register, three lines of defence, vendor diligence, incident response
• Living regulations - the EU AI Act in practice, the US patchwork, cross-border conflicts
• Hard questions you cannot avoid - worker dignity, whose values, the Zero Harm baseline

Anchored throughout in real cases (Moffatt v. Air Canada, COMPAS, SyRI, Robodebt, Mobley v. Workday, SCHUFA, the Italian Garante orders, the New York Times litigation) and the regulations the reader is actually subject to: EU AI Act, NIST AI RMF, ISO 42001, GDPR Article 22, Colorado AI Act, NYC Local Law 144.

The book closes with a ninety-day governance starter programme for an organisation beginning from nothing, plus seven working appendices: AI risk register, impact assessment, model card, vendor due-diligence checklist, incident response playbook, board oversight charter, and a glossary of working definitions.

For the reader who wants compliance built before the regulator asks.

Sotiris Spyrou advises boards and executive teams on AI strategy, transformation, and responsible AI.