GREEN AI AND AI ACT: PIONEERING LEGISLATION OR MERELY AN ENVIRONMENTAL STATEMENT?

Abstract

This article examines how the European Union’s environmental perspective on artificial intelligence has evolved, contrasting the high expectations set forth by the European Green Deal with the subsequent AI Act. Although the European Green Deal acknowledged AI’s potential to enhance energy efficiency, it did not thoroughly address water and energy consumption or the handling of e-waste arising from the development of large deep learning models. Subsequently, the AI White Paper delved more deeply into the environmental dimension of this technology. However, the AI Act—enacted in 2024—does not fully translate these proposals into concrete obligations. The article discusses the main provisions of the AI Act that refer to sustainability, underscoring the absence of direct mechanisms to limit energy consumption, mitigate water footprints, or ensure proper electronic waste management. In this context, two de lege ferenda measures are proposed to remedy these shortcomings: the mandatory inclusion of environmental impact factors in competitive AI benchmarks, and the implementation of an environmental labeling system that informs consumers about the sustainability of the data centers where models operate.