On 10 October, the European Union Agency for Fundamental Rights organised an expert meeting to discuss the preliminary results of the FRA project on Assessing high-risk AI, with a focus on developing guidance for a Fundamental Rights Impact Assessment (FRIA) under the AI Act. The FRA project on Assessing high-risk AI, which started in January 2024, will provide empirical analysis and guidance on how to assess high-risk AI in relation to fundamental rights.
Professor Mantelero was one of the invited experts and his main contribution was on the design of the FRIA template, stressing the importance of an expert-based assessment and the need to streamline the assessment process, avoiding a long list of questions for each potentially affected right.
In this respect, he emphasised that the FRIA is a contextual analysis, so we need to formulate different questions in different contexts, taking into account the potentially affected persons, the context of use and the specificity of the AI used. While it is possible to formulate certain standard questions, the risk of having a questionnaire for non-experts is critical.
Furthermore, he said: “To be comprehensive (is that possible?), we would have to translate into questions all the possible ways in which AI systems could affect a given right”. As a result, the questionnaire will be extremely long. On the other hand, if we focus on just a few key questions, non-experts will think that answering them will complete their task, leaving aside any issues not covered by those questions.
Based on these considerations, Prof. Mantelero promoted a model template focused on the methodology of risk assessment and outlining the key elements and criteria for impact quantification, rather than an automated toolkit focused on an extensive set of general questions for each fundamental right, and with a limited contribution in terms of methodology for assessing the level of impact.