Europe's AI and its environmental cost: SUSTAIN-AI-REG at the forum Civica preview

How can rules, rights and innovation help Europe bridge its divides in artificial intelligence? This was the question at the heart of the round table "L'Europa dell'IA. Regole, diritti e innovazione per superare i divari" (Europe's AI: rules, rights and innovation to bridge the divides). It was hosted in the Sala Europa of the European Parliament Office in Italy. The event previewed the fourth edition of forum Civica, "Oltre i divari" (Beyond the divides), organised by Next Generation EuroPA. The EUI School of Transnational Governance was scientific partner and ANSA media partner, under the patronage of the European Movement Italy.

Taking part in the debate were Sergio Bellucci (UN-mandated University for Peace), Stefano Epifani (Digital Sustainability Foundation), Antonella Fancello (FormezPA and ANCI) and Francesco Rossi Salvemini (Representation of the European Commission in Italy). The Jean Monnet Centre of Excellence SUSTAIN-AI-REG of LUMSA University also contributed, through its scientific coordinator, Luca Megale.

The Centre brought to the debate the issue at the core of its research: the environmental sustainability of AI.

In the Centre's analysis, the regulatory framework is not up to the challenge. On the environmental front, the AI Act was a missed opportunity, and the Digital Omnibus has not changed that. There is more movement in the rules on data centres. Yet more sustainable solutions across the digital value chain already exist:

• smaller, optimised models;

• computing shifted to where and when energy is clean;

• data centres that recover heat and save water.

  
  

They remain in the background because, today, they pay off less. The Centre's proposal is regulation that makes them "the most convenient choice, through public procurement, standards and grid-connection criteria".

Public administrations are the Centre's field of research, and for them there are two keys. The first is proportionality: using AI only when it is needed, with models suited to the task. The second is the reuse of open solutions across public bodies, which cuts costs and consumption and strengthens data sovereignty.