D3.7 - Finalized set of ELSA lab services & report on their performance

This deliverable D3.7 presents the finalised catalogue of ELSA (Ethical, Legal, and Social Aspects) services developed within the AgrifoodTEF project and evaluates their operational performance during the validation phase. The scope of the document encompasses a detailed description of ten specific services provided by European partners, ranging from hazard analysis to ethical scanning and legal guidance, all designed to support the agri-food sector in responsible artificial intelligence (AI) development and navigating complex regulatory environments. Beyond merely listing these services, the report synthesises empirical data collected through customer feedback, national stakeholder workshops, internal cross-service reflections, and developer self-assessments to determine their actual utility and performance. By merging these external market signals with internal operational insights, the report aims to validate the transition of the ELSA infrastructure from a theoretical pilot phase into a robust, industrial-grade ecosystem capable of de-risking agricultural AI.

The primary objective of this deliverable is to assess the performance of the TEF ELSA services in transitioning the ELSA infrastructure from a theoretical framework into a validated operational reality that supports the market introduction of trustworthy AI and robotics in Europe. By analysing the performance data, this report aims to demonstrate how these ELSA services effectively bridge the gap between laboratory prototypes and market deployment for customers, ensuring that technological innovations are ethically sound, legally compliant, and socially acceptable. Furthermore, the document identifies structural challenges and strategic pathways for the future of these ELSA services, aiming to establish a sustainable, long-term service ecosystem that assists stakeholders in the farm-to-fork chain.

The following findings, representing preliminary evidence, are more related to relatively more robust and mature ELSA services, and therefore, they may not be interpreted as statistically generalisable performance metrics for the entire portfolio. Still, new ELSA services can benefit from the results to improve their potential performance.