AGI for healthcare? A call for conceptual discipline in the European Union

The European Commission has recently launched a call for proposals on the topic of artificial general intelligence (AGI) for healthcare. In this editorial we discuss possible interpretations and implications of such a call. It explores the different meanings and discourses around AGI and concludes that it is a contested term that may be linked to significant risks and harms. We then look at how the EU has traditionally dealt with significant but unquantifiable risks using the precautionary principle and how it current aims to deal with risks of AI, focusing on the EU AI Act. We conclude that the use of the term AGI in the call is fundamentally problematic. It may stand for advanced AI technology, in which case there is no need to use such a contested term. Alternatively, it refers to qualitatively novel types of AI technologies with unpredictable capabilities. If this is the case, then it may raise significant risks and potential harms and funding work on it conflicts with the long-standing precautionary approach to such risks. Either way, we believe that increased attention to conceptual discipline would benefit the EU’s Horizon Europe research framework programme and European AI policy more broadly.