ELI hosted a highly anticipated webinar on its Response to the European Commission's guidelines on the application of the AI Act, focusing on the definition of an AI system.
The webinar was opened by Sir Geoffrey Vos (ELI Vice-President; Master of the Rolls and Head of Civil Justice in England and Wales), who welcomed participants and emphasised the importance of defining AI systems precisely, stating that ambiguity could lead to unintended consequences in both legal and practical applications. Sir Geoffrey also acknowledged the need for continued dialogue to refine these definitions as AI technology evolves.
Christiane Wendehorst (ELI’s Scientific Director; Professor at the University of Vienna) and Bernhard Nessler (Research Manager at the Software Competence Center Hagenberg (SCCH); Vice-President of the Austrian Society for Artificial Intelligence) jointly presented the Three-Factor Approach developed by ELI as part of its Response to the consultation (available here) to delineate AI systems for the purpose of the AI Act. They explained that this approach evaluates systems based on three key dimensions:
Factor I: The amount of data or domain-specific knowledge that went into development.
Factor II: The extent to which know-how is created during operation.
Factor III: The degree of formal indeterminacy of outputs.
This approach, they stressed, is crucial for regulatory clarity and the consistent application of the AI Act across various sectors. Among other things, their presentation also explored practical challenges in differentiating between complex AI systems and more straightforward IT tools, such as spelling and grammar checkers, which could be misclassified as AI systems under the current guidelines.
Karine Perset (Acting Head of the OECD Division on AI and Emerging Digital Technologies) gave the OECD’s perspective on AI regulation, noting that international collaboration is key to setting global standards for AI. Perset highlighted the OECD’s work in creating common principles for AI governance and stressed the need for clarity in defining high-risk AI systems, especially as AI’s role in decision-making processes grows. Among other things, she called for further refinement of the definition in future consultations to better address the ethical and societal impacts of AI technologies.
Tatjana Evas (Legal and Policy Officer, European Commission) provided insight into the European Commission’s recent publication of guidelines on AI system definition, released on 6 February 2025. She emphasised that these guidelines aim to facilitate the implementation of the first rules under the AI Act, particularly regarding high-risk AI systems. Evas noted that ongoing feedback from stakeholders will continue to inform the regulatory framework and that the Commission remains committed to refining the guidelines as the AI landscape evolves.
The webinar concluded with an engaging Q&A session, during which participants expressed concerns over the practical application of AI definitions in industries such as healthcare and autonomous driving, where AI technologies have significant implications for safety and accountability.
The session highlighted the ongoing challenges in AI regulation, particularly in defining AI systems clearly and consistently. It also underscored the need for future updates to the AI Act to keep pace with technological advancements and provide clearer guidance for businesses and policymakers alike.
The recording of the webinar is available below.
