Nicoletta Rangone: Reflections on an innovation capable of protecting the environment and citizens' fundamental rights (automatically translated from Italian)

On the occasion of the conference "Improving Administrative Decisions using AI, Experimentation and Sandboxes," organized by the Universitat Oberta de Catalunya and the Universitat de Barcelona , Nicoletta Rangone (Director of the Center) was interviewed by Spanish journalist Saül Gordillo for the magazine Algoritmo Trasparente . The interview is available in video and text format, in Spanish , English , Catalan , and Italian . During this conversation, numerous ideas emerged for regulation that is favorable to innovation, while also protecting the environment and citizens' fundamental rights. Some excerpts are summarized below, with reference to the original text for further information.

AI and environmental sustainability: an open problem

During the interview, the professor challenges, at several points, a number of widespread assumptions about the mutually exclusive relationship between technological progress, the environment, and fundamental rights. While the urgency of reconciling the drive for rapid innovation with rights is evident, the European regulator's goal should not be to favor one of the two needs while hindering the other, but to balance them with the greatest possible awareness.

Achieving this could certainly be facilitated by defining a shared method for measuring the environmental impact of AI systems . Currently, it is unclear what should be assessed, when a system's life cycle begins and where it ends. This aspect strongly affects the collection of comparable data. In the absence of transparency that also extends to the environmental impact of AI infrastructures, models, and systems, it is impossible for citizens, administrations, and businesses to make informed choices to limit their environmental footprint. While waiting for this outcome, public and private entities could nevertheless make a series of choices that move toward a lower environmental impact of AI, for example: reusing open source models , because they avoid or reduce the impact of training activities; using Small Language Models when greater computational resources are not necessary; adopting system development or procurement criteria based on environmental standards (in terms of water or energy consumption); the definition of development or purchasing criteria because it is at this stage that compliance with certain environmental standards (for example, in terms of water or energy consumption) and transparency of the measured results may be required.

This need for clarity and shared awareness has transcended the confines of legal and technological reflection, finding a strong appeal even beyond secular institutions. In particular, Leo XIV's encyclical Magnifica Humanitas addresses the issue head-on, calling for collective accountability for the impacts of AI in order to exploit its potential without compromising our future.

European AI Regulation: Strengths, Weaknesses, and Risks to Effectiveness

Responding to the narrative that European regulation tends to paralyze innovation processes, Nicoletta Rangone clarifies several key elements for analyzing the phenomenon. The AI Act clearly places the protection of citizens' fundamental rights at its core. The aim is to strike a delicate balance between innovation, rights, and competition, within the framework of a human-centric model. This approach should be a source of pride for European citizens.

What fuels the false dilemma between innovation and rights is viewing the rights-driven approach as an obstacle, rather than a prerequisite, for technological development. At the same time, the professor recognizes a weakness of the European Regulation: its enormous complexity. Despite setting important goals, the AI Act, in its original version, included 113 articles, divided into 13 chapters, 180 recitals, and 13 annexes. This structure is complemented by delegated and implementing acts, harmonized standards, codes of best practice, and guidelines: instruments of varying nature and legal force that contribute to defining a massive regulatory architecture.

All this is combined with other relevant interventions for the digital ecosystem: not only GDPR , DSA and DMA , but also new initiatives, including the proposed Cloud and AI Development Act ,Chips Act 2.0 , included in the Tech Sovereignty Package , and the Industrial Accelerator Act . The complexity of this framework can weigh down the compliance efforts of businesses and public administrations and undoubtedly requires simplification, but certainly not deregulation.

However, simplification also presents risks. The first is that it can generate legal instability when it repeatedly modifies deadlines and the content of obligations. The Digital Omnibus on AI , now Regulation (EU) 2026/1744, postponed the application of various obligations for high-risk systems (for example, the regulation of high-risk systems listed in Annex III of the EU AI Act has been postponed to 2 December 2027, and to 2 August 2028 for those considered high-risk if they are a safety component of a product regulated by EU legislation listed in Annex I of the EU AI Act). The reform has modified the provisions of the EU AI Act, simplifying certain registration requirements and reshaping the AI literacy requirement (supportive measures are required, but there is no requirement to ensure a specific individual level). In parallel, the general Digital Omnibus proposal, COM(2025) 837 , still under examination by the co-legislators, also addresses the protection of personal data, including a derogation for the residual processing of special categories of personal data in the development and operation of AI systems and models subject to conditions and guarantees (it is therefore not a general authorisation to use sensitive data for training as feared).

Overall, the frequent close changes to the rules and related deadlines can increase compliance costs and encourage (in businesses and administrations) a wait-and-see attitude with respect to the investments needed to ensure compliance.

Regarding the method used, the simplifications in digital ecosystem regulation were achieved through the two aforementioned Omnibus initiatives (which are part of the 12 Omnibus initiatives adopted between January 2025 and June 2026). These reforms reveal a worrying trend: the European Commission is progressively scaling back the decision-making model based on broad stakeholder participation and solid empirical evidence. The Omnibus decision-making process has often involved consultations aimed solely at experts, without involving citizens, and without a prior impact assessment. This, moreover, not only reduces the guarantees of participation, but also limits the information needed to make evidence-based decisions. The Commission communication of 28 April 2026 ( A Simpler, Clearer and Better Enforced EU Rulebook , COM(2026) 380 ), also presents elements that should be carefully observed, allowing for flexibility in consultations and procedural derogations in exceptional situations of urgency. The risk is that the pursuit of speed in regulatory processes compresses the spaces for participation and impoverishes the information base available to public decision-makers. Urgency is often inevitable in times marked by crises of various kinds, but it becomes problematic when it permanently replaces listening, impact analysis and transparency.

What do the experiences of Spain and France tell us?

The event highlighted how various Spanish local governments, while operating within the scope of European data protection legislation, have encouraged and supported the development of innovative projects for cities by making data collected by the cities themselves available. This approach runs counter to the more conservative approach to GDPR implementation that seems to characterize other countries, including Italy. Among the applications presented are the Barcelona experience, which uses AI to detect illegal tourist accommodations , a public order issue, but also a problem for citizens' access to housing – as highlighted by a recent European proposal on the matter; and the management of the sandboxes in Madrid , Barcelona , and Valencia .

The conversation also touches on the topic of digital sovereignty, starting with the case of France, which has decided to invest in national digital infrastructures like the cloud and in companies like Mistral, and has embraced a " frugal AI " approach. At the same time, the country has chosen to limit its dependence on certain foreign platforms in the most sensitive sectors. This also offers a lesson for Italian public administrations to evaluate where their data is stored, what legal system protects it, and what level of technological dependence certain decisions generate. All of this is done with the understanding that digital sovereignty measures should not lead to nationalistic closures, but rather ensure that the technologies used are consistent with the values and sensitivities of the countries that use them.