Professor Mantelero participated in a panel organised by the European Commission, Research Ethics and Integrity Sector, at the EuroScience Open Forum in Katowice. As part of his ongoing work on the Guidance Framework on Fundamental Rights Impact Assessment in AI Research Assessment for the European Commission, which looks at future EU-funded projects in AI and the use of AI, Prof. Mantelero highlighted some key points on the interplay between the AI Act, research and the ethical assessment of EU-funded projects:
📌 Given the wide range of uses of AI and its potential impact on fundamental rights and freedoms, ethical assessment should be complemented with FRIA, considering the specificity of the latter and associated methodologies.
📌 The FRIA is a values-oriented risk assessment, which means that fundamental rights cannot be overridden by any kind of competing interest.
📌 The contextual nature of the FRIA requires more careful evaluation of the possible scenarios of use and the associated configuration of key AI system features.
📌 The FRIA requires the combination of fundamental rights expertise with the scientific domain knowledge relevant to each project, in order to assess and manage potential impacts.
📌 The AI Act provides a research exception that refers to AI systems and models “specifically developed and put into service” for scientific research and development, when scientific research and development is the sole purpose of AI (Article 2.6 AI Act).
📌 Some of the research on AI that is not covered by this exception is part of the broader exception for research and development activities (Article 2.8 AI Act).
📌 The European Commission and EU research bodies should therefore complement the AI Act with regard to both ethical principles and fundamental rights.
📌 Regardless of the scope of the FRIA in the AI Act, the protection of fundamental rights recognised by the EU legal framework requires the existing ethical assessment and review procedure to be strengthened by the introduction of a specific component focusing on fundamental rights.
📌 Finally, general obligations to protect fundamental rights are also relevant in the case of AI applications for research purposes that are not regulated by the AI Act and for applications that do not pose a high risk.