How to Assess Trustworthy AI in Practice

Assessment requires examining AI systems across their entire lifecycle, not just at deployment. The Z-Inspection® process evaluates trustworthiness at different stages, recognizing that ethical issues can emerge at various points from development through operation.

Ethical tensions are inevitable and must be actively discussed. Rather than assuming AI systems can be perfectly ethical, the methodology emphasizes identifying and deliberating on competing values and trade-offs through socio-technical scenarios—a practical approach for teams to navigate real-world complexity.

The EU's trustworthy AI guidelines can be operationalized in practice. This report demonstrates that abstract principles from the EU High-Level Expert Group can be translated into concrete assessment procedures, making trustworthy AI evaluation actionable for both researchers and practitioners rather than remaining theoretical.

Healthcare provides important lessons for trustworthy AI assessment. The independent assessments conducted in healthcare settings revealed practical challenges and solutions relevant to other high-stakes domains, suggesting that sector-specific experience can inform broader responsible AI practices.

Rigorous assessment requires systematic, structured processes. The Z-Inspection® methodology emphasizes that trustworthy AI evaluation should follow a holistic, methodical approach rather than ad-hoc reviews, helping organizations conduct consistent and comprehensive trustworthiness checks throughout an AI system's lifetime.