A Framework for the Categorisation of General-Purpose AI Models under the EU AI Act

Lorenzo Pacchiardi, John Burden, Fernando Martínez-Plumed, Jose Hernandez-Orallo, Emilia Gomez, David Fernández-Llorca

Keywords: EU AI Act, General-Purpose AI Models, capabilities, generality, ADeLe, measurement scales, AI evaluation, AI policy, AI regulation, cognitive domains

TL;DR: We introduce a capability-based framework using measurement scales for categorising General-Purpose AI Models under the EU AI Act

Abstract: We propose a framework for categorising AI models as General-Purpose AI (GPAI) models, based on their capabilities and generality, as defined in the European Union (EU) AI Act. Our framework breaks down the core components of the GPAI definition into measurable elements, focusing on four primary cognitive domains: Attention and Scan, Comprehension and Compositional Expression, Conceptualisation, Learning and Abstraction, and Quantitative and Logical Reasoning. We suggest using the Annotated Demand Levels (ADeLe) procedure to evaluate AI models’ capabilities in these domains, and provide a methodology for combining domain-level scores into a single measure of generality. The framework is illustrated with empirical results from existing models, and policy recommendations are made for selecting thresholds and metrics for GPAI categorisation.

Submission Number: 39