Lessons Learned in Performing a Trustworthy AI and Fundamental Rights Assessment

Assessment frameworks can be practically combined: The pilot successfully integrated two different evaluation approaches—the Z-Inspection® Trustworthy AI process and the EU's Trustworthy AI framework—alongside a Fundamental Rights Impact Assessment (FRAIA), demonstrating that multiple responsible AI assessment methods can work together in real-world government applications.

Environmental AI applications require careful rights evaluation: Even beneficial uses of AI like satellite-based heathland monitoring for nature reserve management need systematic assessment of fundamental rights impacts, suggesting that responsible AI practices should apply across all government sectors, not just high-risk applications.

Practical experience reveals implementation gaps: The report emphasizes sharing "lessons learned" from conducting these assessments, indicating that while frameworks exist for trustworthy AI, their real-world application by government agencies revealed challenges and insights that weren't apparent in theory alone.

Collaboration between government and inspection bodies is valuable: The pilot involved cooperation between provincial authorities, national government (Ministry of the Interior), and the Z-Inspection Initiative, suggesting that responsible AI implementation benefits from multi-stakeholder involvement and specialized expertise beyond individual agencies.

Satellite imagery and machine learning in environmental monitoring needs governance: The pilot focused on an increasingly common practice—using deep learning with satellite data for environmental decisions—highlighting the need for trustworthy AI assessments as these technologies become standard tools in government environmental monitoring.