Hostname: page-component-77f85d65b8-grvzd Total loading time: 0 Render date: 2026-03-28T06:30:00.941Z Has data issue: false hasContentIssue false
  • English
  • Français

Operationalising the Twin Transition: From EU Constitutional Principles to National AI Strategies

Published online by Cambridge University Press:  24 March 2026

Alberto Quintavalla Open the ORCID record for Alberto Quintavalla [Opens in a new window] ,
Edoardo Celeste  and
Alba Perez Victorio
Alberto Quintavalla*
Affiliation:
Erasmus School of Law, Erasmus University Rotterdam, Netherlands
Edoardo Celeste
Affiliation:
School of Law and Government, Dublin City University, Ireland ADAPT Centre, Ireland
Alba Perez Victorio
Affiliation:
Law and Tech Research Cluster, Dublin City University, Ireland European University Institute, Italy
*
Corresponding author: Alberto Quintavalla; Email: quintavalla@law.eur.nl
Rights & Permissions [Opens in a new window]

Abstract

EU policy documents often present the green and digital transitions as “twinned” and mutually beneficial. However, it is now widely acknowledged that new technologies, such as AI, have significant environmental costs. This awareness has induced the EU to recalibrate its approach to the twin transition and prompted the elaboration of new soft-law principles of digital sustainability, an early form of “eco-digital constitutionalism” at EU level. This paper examines the national AI strategies of the twenty-seven Member States to assess the extent to which this revised approach of the EU is reflected at the national level. Our analysis reveals that the strategies in question often differ starkly in the manner in which they prioritise the various matters that pertain to the intersection of environment and AI. We argue that these inconsistencies, which are rooted in different attitudes to the economics of innovation, cannot be solved fully through soft-law harmonisation.

Keywords

eco-digital constitutionalismnational AI strategiestwin transition

Information

Type
Articles
Information
European Journal of Risk Regulation , First View , pp. 1 - 15
Creative Commons
Creative Common License - CCCreative Common License - BYCreative Common License - NCCreative Common License - ND
This is an Open Access article, distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives licence (https://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided that no alterations are made and the original article is properly cited. The written permission of Cambridge University Press or the rights holder(s) must be obtained prior to any commercial use and/or adaptation of the article.
Copyright
© The Author(s), 2026. Published by Cambridge University Press

I. Introduction

Contemporary EU policy revolves around the idea of twinning the digital and the green transitions. At first sight, the digital transition suggests itself as a useful driver of its green sibling: new technologies can enhance resource efficiency and decrease greenhouse gas emissions from physical processes. However, this simplistic representation of the relationship between the two transitions omits an important problem, namely the adverse environmental impact of digitalisation. In fact, the scientific literature acknowledges that new technologies can sometimes contribute to environmental degradation – the development of AI systems demands substantial energy, water and raw materials, generating a negative ecological impact.Footnote 1

In this scientific and intellectual milieu, which is no longer shaped by the techno-optimism of yesteryear but by the more immediate prospect of hard bargains and uncertain trade-offs between green and digital, the EU has introduced soft-law “digital sustainability” principles as an attempt to operationalise the twin transition and bypass its core dichotomic tension.Footnote 2 Building on previous EU efforts to enhance environmental protection and safeguard the digital world, these principles represent a first step to lay the constitutional foundations for a successful twin transition.

The existing body of legal scholarship has extensively examined the capacity of the EU to integrate its digital and environmental governance frameworks effectively. For example, a critical analysis has been conducted on the core public values that underlie the approach of the EU to the twin transition, revealing the normative priorities that influence policy decisions.Footnote 3 Others have questioned whether the current EU legal framework is adequately equipped to manage the complexities that are inherent in the twin transition,Footnote 4 or have proposed reforms to the legal architecture of the Union to address these intertwined challenges more effectively.Footnote 5 Furthermore, research has explored the degree to which the EU has incorporated environmental concernsFootnote 6 or adhered to broader sustainability principles.Footnote 7

This scholarship has focused predominantly on the highest level of governance, that of the EU. Any evaluation of twin-transition policy that is restricted in this fashion is bound to be misleading. The Member States continue to hold a decisive influence on the twin transition in light of the principle of subsidiarity in EU law. This paper contributes to exploring such a division of responsibility and its repercussions for the dual impact of technology on the environment. We do so by studying the constitutionalisation efforts of digital sustainability principles at EU level and assessing to what extent they are operationalised in the twenty-seven national agendas on the most critical technological development of our day, namely AI. We argue that, despite an incipient “eco-digital” constitutional framework laid down by the EU, Member States’ implementation tends to fragmentation and discordancy. The paper posits that, on top of the soft-law nature of the constitutional foundations of the twin transition that have been articulated by the EU, this result is also due to the wide divergence in national views on the import of ecological sustainability for AI and the current geopolitical scenario in which the EU operates.

The remainder of this article is structured in five parts. Section II explains how EU efforts to constitutionalise environmental and digital rights have so far surfaced as two distinct and siloed responses to contemporary societal challenges. Section III illustrates how the emergence of digital sustainability principles in the Declaration on Digital Rights and Principles may be regarded as a first attempt to lay down the constitutional foundations of the EU twin transition in a soft-law instrument. Section IV then explains why any representation of the dilemma of regulating the twin transition which focuses exclusively on EU level is incomplete, with a particularly strong emphasis on the concept of subsidiarity and how it impacts the national operationalisation of EU digital sustainability principles in the domain of AI. Section V contains the results of our analysis of twenty-seven national agendas on AI, and it inquires why, despite recent EU soft-law constitutionalisation efforts, actual policy remains profoundly fragmented and largely uncoordinated. In Section VI, we conclude by positioning the current asynchrony between EU and Member States’ operationalisations of twin transition objectives in the AI field in the current geopolitical scenario.

II. Siloed responses: environmental and digital constitutionalism in the EU

The twin transition is defined by the coexistence of two interconnected domains: environmental protection and digitalisation. Historically, these domains have operated in isolation, with separate policy frameworks and regulatory efforts addressing each area independently.Footnote 8 Environmental protection initiatives have predominantly aimed at safeguarding natural resources, promoting sustainability, and mitigating the impacts of climate change. In contrast, digitalisation has been driven by technological advancements and the integration of digital tools across various sectors of society. This separation has resulted in the development of distinct constitutional responses, respectively the expression of what the scholarship called environmental and digital constitutionalism.

The idea of environmental constitutionalism emerged as a response to the need to adapt constitutions to the negative influence that industrial economies exert on the environment.Footnote 9 The origins of this notion can be traced to the incorporation of provisions on environmental protection into various US state constitutions.Footnote 10 Environmental constitutionalism is also often associated with the environmental movements of the 1960s and 1970s and with the UN Stockholm Declaration of 1972, which emphasised environmental protection at the global level.Footnote 11 The Robinson Township case, which was decided by the Pennsylvania Supreme Court in 2013,Footnote 12 did much to reinvigorate this field of legal scholarship,Footnote 13 as did the work of numerous Latin American scholars from the end of the 2000s, which built on contemporaneous constitutional reforms in that part of the world.Footnote 14 At present, environmental protection is covered extensively by constitutions around the world, and the academic debate about the issue appears to have stagnated somewhat as a result of its past successes.Footnote 15

At EU level, environmental protection was not couched in the language of Treaty rights until 2000, which saw the promulgation of the Charter of Fundamental Rights.Footnote 16 The Charter established that EU policies must ensure “a high level of environmental protection” and promote improvements in environmental quality.Footnote 17 The CJEU has developed sophisticated doctrines on environmental matters and striven to ensure compliance with EU environmental legislation.Footnote 18 Individual Member States began to incorporate provisions on environmental protection into their constitutions during the second half of the 1970s.Footnote 19 Although approaches and levels of commitment varied,Footnote 20 the constitutions of Greece (1975), Portugal (1976) and Spain (1978) were all drafted to integrate environmental considerations. Several national constitutions have since been revised to include rules on the protection of the environment,Footnote 21 one of the most recent examples being the “environmental reform” of the Italian constitution of 2022.Footnote 22 Perhaps the most striking illustration of the ascent of environmental constitutionalism to date is the Charter for the Environment, an appendix to the French Constitution introduced in 2005, which is now held to possess constitutional status.Footnote 23

The emergence of environmental constitutionalism stemmed from the need for the constitutional ecosystem to react to the adverse effects of contemporary society on the environment. A similar constitutional response, what the scholarship called “digital constitutionalism,” emerged to address the ills and promises of the digital transformation.Footnote 24 Both environmental and digital constitutionalism do not aim to revolutionise existing constitutional theory, but rather to help interpret ongoing processes of constitutionalisation of rights. Environmental and digital constitutionalism do not depart from contemporary constitutionalism but seek to adapt it to social change. This is particularly relevant in the digital context as technological innovation and, particularly the rise of AI, has yielded both positive and negative effects on the protection of fundamental rights.Footnote 25 For example, digital technologies have made it easier to exercise the right to freedom of expression and speech-based rights, such as freedom of religion or assembly, more generally. At the same time, new technologies have aggravated the risks of surveillance by state actors and created new threats to offline and online rights.

Scholars have only recently begun to pay attention to digital constitutionalism.Footnote 26 The first generation of digital constitutionalist scholarship analysed non-traditional constitutional sources, such as non-legally binding digital charters and social media’s terms of service. EU constitutional responses to the challenges of the digital revolution started to garner academic attention in the second iteration of that scholarship.Footnote 27 The Court of Justice of the EU has played a crucial role in the constitutionalisation of digital rights. During periods when EU law struggled to offer clear solutions to the challenges posed by technological advancements, the Court provided an evolving interpretation of existing legislation.Footnote 28

The response of the EU legislator came later, but with a significant intensity. The scholarship identified a phenomenon of “actification” in the digital field.Footnote 29 Key pieces of EU digital law are indeed known as “acts” and they directly refer to their subject matter. To mention a couple of examples, Regulation (EU) 2022/2065 is more simply called Digital Services Act, while Regulation (EU) 2022/1925 is known as the Digital Markets Act. As a technical matter, EU law does not recognise “acts” as distinct legal sources; yet, the scholarship has interpreted this use as a means to raise public awareness on key pieces of legislation which are part of a coherent regulatory programme aiming at tackling technological advancements.Footnote 30 This is what Floridi called the “hexagram of digital constitutionalism,” referring to the GDPR, the DSA, the DMA, the Data Governance Act, the EHDS Regulation and the AI Act.Footnote 31 Despite the EU’s robust regulatory response to the digital revolution, technological innovation continues to advance rapidly. Consequently, EU Courts must tackle the gaps left by legislation, and the significance of soft law at EU level should not be overlooked.

The constitutionalisation of environmental and digital rights has largely developed in isolation at EU level. However, a notable exception to this lack of an integrated perspective lies in a soft-law document: the Declaration on Digital Rights and Principles adopted by the EU in 2023.Footnote 32 This Declaration, which will be discussed in more detail in the following section, incorporates principles that acknowledge the interconnectedness of environmental and digital challenges. This represents a significant milestone in what can be denoted as an effort of “eco-digital constitutionalism” at EU level, as it marks one of the initial steps toward establishing the constitutional foundations of the effective twinning of the digital and green transition.Footnote 33 These “eco-digital principles” can offer guidance to both national and EU policymakers. Additionally, they can serve as a platform to reconnect and further develop legislative provisions related to environmental and digital issues that have historically been fragmented across various laws.

Here, one should, however, note that the concept of “eco-digital constitutionalism” is not intended to establish a new normative framework that advocate for the integration of digital and environmental concerns within the EU. Instead, it serves primarily as an analytical tool that is designed to identify the emergence and subsequent convergence of policy and regulatory provisions that address the intersection of digitalisation and ecological sustainability. Employing this framework makes it easier to gauge the extent to which these two dimensions are increasingly intertwined within the constitutional and regulatory landscape of the EU. At the same time, our analytical perspective sheds light on the potential tensions that could arise between the digital and environmental domains by offering insights into the ways in which the policy framework of the EU and, as we will demonstrate later, the follow-up responses of the Member States are evolving to manage these conflicts. In this way, our analysis is a more nuanced approach to the ongoing transformation of policy and law in both the EU and its Member States, whereby sustainability and digitalisation are no longer treated as isolated spheres. Instead, they are increasingly recognised as interdependent elements of a comprehensive vision that seeks to integrate technological progress with environmental sustainability goals.

III. The digital and the green transition in the EU soft law: from synergies to tensions

The previous subsection showed that constitutional answers addressing environmental protection, on the one hand, and the risks emerging from digitalisation, on the other hand, are well-established at EU level. However, the rapprochement of these two dimensions has only started in recent times. The general idea of a twin transition began gaining traction after the President of the European Commission used this term in a speech at the European Parliament in 2019.Footnote 34 One of the first Communications to refer to it was “The European Green Deal,” which anticipated that the twin transition would shortly form the core of a new industrial strategy for the EU.Footnote 35 A year later, the Commission outlined the various means by which the relationship between technological innovation and environmental protection could become mutually beneficial and reinforcing.Footnote 36 The Council likewise underlined the need for coherent policy and close cooperation in the two policy domains.Footnote 37 A certain vision of positive feedback loops as the conceptual essence of the twin transition then took hold.Footnote 38 As a result, much EU policy in that period was premised on positive assumptions, such as that digital technology and data always conduce to more accurate climate and environmental monitoring and prediction, to more efficient resource allocation, to cheaper and more efficacious regulatory compliance.

This enthusiasm has more recently been tempered considerably by the recognition that the digital and the environmental may also come into tension, and that it is sometimes necessary to strike difficult and uncertain trade-offs between the two. For example, a Report of the Special Committee on Artificial Intelligence in a Digital Age that was commissioned by the European Parliament stressed the point that the uptake of modern technology is liable to aggravate problems in both human society and nature, which obviously runs counter to the environmental agenda of the Union.Footnote 39 These concerns were promptly integrated into EU policymaking: the Commission now speaks openly of the tensions between the two transitions and not only of the synergies between them.Footnote 40

This nuanced perspective on the relationship between environment and digitalisation is also informed by principles adopted by the EU, which can offer the constitutional foundations to operationalise the twin transition. The EU Declaration on Digital Rights and Principles to which we referred in the preceding section is a case in point. It contains “digital sustainability principles,” as well as several values and fundamental rights that the EU perceives as essential to the formation of a good digital environment. The principles in question revolve around empowerment, fairness and social inclusion. The Declaration is also intended to enhance the digital sovereignty of the Union in a globalised world and to promote the “European way for the digital transformation.”Footnote 41 Thus, although the Declaration formulates core digital rights and principles for the EU, it is also meant to guide non-EU policymakers.Footnote 42 It comprises six chapters, of which the last is dedicated to the sustainability of digital technologies.

Chapter VI contains three distinct digital sustainability principles.Footnote 43 There is an overarching “principle of sustainable development, deployment and use of digital technologies,” whose content is scattered across the various political commitments listed in the chapter. It has two facets: the promotion of digital technologies that have minimal negative consequences for the environment and society, that is, the management of the tensions between the two transitions, and the encouragement of the use of digital technologies in the pursuit of positive environmental outcomes, that is, the realisation of the synergies between the digital and the green. Instead, the other two more specific principles refer, respectively, to the sustainability of digital products and services, and to the disclosure of their environmental impact.

The recognition of these principles in the Declaration evidently strengthens the EU constitutional foundations of the twin transition, in those ideas that had previously only been ventilated in some lesser-known policy instruments are now enshrined in a solemn document which bears the signatures of the Presidents of the Council, the Commission and the European Parliament.Footnote 44 At the same time, it remains the case that the Declaration has no binding legal force – nobody can sue on any of the rights that it contains, and nobody can ask a court or an administrative authority to enforce them; at most, the three principles can serve as interpretative props in borderline cases about laws that actually bind. Its purposes are thus largely declarative and didactic.

In fact, whether the darker aspects of the twin transition have had any palpable influence on hard EU law is very much open to question – both the twinning of the transitions and the tension between their objectives are largely unaddressed in any specific piece of binding EU legislation.Footnote 45 The paradoxes of twinning are also only ever noted in documents which are incapable of direct legal application, of which the Declaration is one, and even those cursory mentions do not appear to afford exhaustive solutions to the relationship between the digital and the green.Footnote 46 Furthermore, the role of the environment in EU digital regulation is decidedly secondary. The one exception to this proposition is the AI Act, and it is the kind of exception that serves principally to confirm the rule – environmental concerns were only acknowledged in the AI Act after a last-minute intervention by the European Parliament.Footnote 47 Likewise, digitalisation is also almost never addressed directly in EU environmental law. When it is addressed, the references to it tend to be brief and the regulatory approach tends to be increasingly procedural, relying on private actors.Footnote 48 For instance, the recast Energy Efficiency Directive mandates the establishment of a monitoring scheme to oversee energy consumption of data centres, but does not offer substantive solutions to harmonise Member State law on this issue.Footnote 49

This allows us to come back to the idea of eco-digital constitutionalism as a propeller of a gradual process that is laying down guiding principles for the twin transition, starting at the level of soft law. The Declaration of Digital Rights and Principles is an attempt to establish a constitutional foundation that enables the EU to offset the intrinsic dichotomy between EU digital and green objectives. However, the Declaration is not legally binding, and its utility lies largely in its conceptual “normative retrofitting” of pre-existing rights and principles in a manner that accounts for the dual impact of digitalisation on the environment.Footnote 50

IV. From the EU to member states: operationalising the twin transition in the AI sector

Under the constitutional framework put in place by the EU, the success or failure of efforts of the twin transition also depends on the active participation and commitment of individual Member States. This is so because, for the most part, the actuation of the principles, regulations and policy agendas that the EU fashions is a matter for the individual Member States. In other words, the attainment of objectives at EU level depends on action at the Member State level, and the fortunes of projects like the twin transition depend on the alignment between the two. As a matter of law, this idea is operationalised in Article 5 of the Treaty on the European Union, which contains the principle of subsidiarity.Footnote 51 The principle, in brief, is that decisions should be made at the most local level possible unless there is evidence that action at that level would be ineffective.

Since the principle of subsidiarity applies to nearly all sub-domains of the twin transition, we will focus on the operationalisation of EU twin transition objectives in the specific context of AI. AI strikes us as a good subject of analysis because, at the time of writing, it is widely considered to be the most powerful socio-economic manifestation of digitalisation and because both its benefits for environmental protection and the harm which it is liable to cause are increasingly being studied in the legal literature.Footnote 52

For AI, subsidiarity means that the EU formulates policy and creates regulatory frameworks, but the Member States shape this guidance into concrete measures. For instance, the EU AI Act urges the Member States to “encourage and facilitate” the development of codes of conduct to ensure that the requirements for high-risk AI systems are applied, on a voluntary basis, to all AI systems.Footnote 53 Even though the EU did not create a universal obligation to draft codes of conduct through the AI Act, the Member States are encouraged to ensure the creation of such codes for all AI systems.Footnote 54

This arrangement is thought to be effective because it allows the Member States to adapt EU policy to their national contexts. Beyond the highly specific example of the codes-of-conduct obligation, the Member States have thus far actuated EU policy on AI by writing and adopting national policies on AI. The European Commission has been instrumental in steering the Member States toward the development and implementation of those policies. In 2018, it released a Coordinated Plan on AI that set a clear framework for the Member States to follow.Footnote 55 Each Member State was urged to formulate national AI strategies that not only align with the overarching objectives of the Union but also foster innovation and competitiveness.Footnote 56 For instance, the 2018 Plan emphasised several goals, including the promotion of ethical AI practices and the creation of a skilled workforce which is capable of leveraging this rapidly evolving technology.Footnote 57

The Plan was revised three years later to include the Proposal for what is now the EU AI Act. The updated Plan focused on four key objectives: establishing enabling conditions for the development and adoption of AI in the EU, positioning the EU as a hub for excellence from research to marketisation, ensuring that AI technologies benefit people, and fostering strategic leadership in high-impact sectors.Footnote 58 Notably, and in contrast to the first Coordinated Plan of 2018, the update refers to the need to balance the benefits of AI for the environment and its environmental footprint expressly.Footnote 59

The discussion of the third objective of the Plan, “Ensure that AI works for people and serves as a force for good in society,” touches on the social and environmental impacts of AI technologies. It recognises the numerous environmental benefits of AI, such as optimal energy use and improved resource management. At the same time, it emphasises sustainability.Footnote 60 The updated Plan identifies the environment as one of the key sectoral areas for the fourth objective, which is to promote strategic leadership in high-impact sectors.Footnote 61 The dual environmental impact of AI is explicitly acknowledged in that objective as well. While AI technologies can facilitate the attainment of the goals of the Green Deal, they also consume much natural resources and energy. The updated Coordinated Plan calls for further research into greener AI solutions and the development of key performance indicators, which are essential for identifying and measuring both the positive and the negative environmental impacts of AI. Furthermore, it outlines key actions that the Member States must take to support the twin transition.Footnote 62 The Member States are encouraged to share insights and effective strategies from their national “green AI” initiatives to build a collective understanding of best practices in environmentally friendly AI development and implementation. The Member States should also integrate green AI into university curricula and AI training programmes. Collaboration with national ICT stakeholders and standardisation bodies is essential for the drafting of deployment guidelines and the creation of standardised assessment methodologies. These frameworks are expected to facilitate the adoption of green AI in infrastructural grids, agriculture and urban development.

Beyond the composition of national agendas, which form the subject matter of the next section, the Member States have engaged in cross-border collaboration for realising the objectives of the twin transition. For example, in 2021, twenty-six Member States, Norway and Iceland signed a Declaration on the twin transition, which includes a commitment to develop and implement energy-efficient algorithms.Footnote 63 Likewise, eighteen Member States have subscribed to a demand to examine the net environmental impact of the digital sector.Footnote 64 As we will show in the next section, several Member States have built on this momentum and revised their pre-existing national AI agendas to account for the dual environmental impact of AI technology. These agendas matter because they reflect the impact of a wide variety of stakeholders, including government agencies, academic institutions, industry, and civil society, and because they contain clear and concrete objectives. Furthermore, they educate the public about the impact of digital technology on the environment (among other matters). Education of this kind has been shown to be essential for the successful uptake of new technology.Footnote 65

V. National AI agendas and the EU twin transition: a fragmented approach

We now turn to the national agendas themselves. We must note at the outset that they are far from uniform in content, purport or coverage. This diversity is the result of differences in political ideology, cultural attitudes to technology, and levels of public engagement with AI. Some countries emphasise the positive impact of AI on the environment, while others focus more sharply on the potential harms. Furthermore, the documents in question were written at different times. The countries that were more cautious and waited longer before publishing their plans, especially those written after the 2021 Communication to which we referred previously, have developed more balanced views on the dual impact of AI technology.

There are twenty-seven separate national AI agendas, corresponding to the number of EU Member States. In order to examine these agendas systematically, we initially relied on the technical report of national strategies that the EU has published.Footnote 66 Our first step involved verifying the currency of each document to ensure that the versions that we included in our analysis were the most recent and that they had been adopted formally by the corresponding Member States. This verification process included cross-referencing official government websites and relevant policy updates to confirm that no subsequent modifications or newer versions had been adopted. In several cases, we had to rely on more recent documents in order to account for temporal differences to the approach of a given Member State to the twin transition.Footnote 67 Furthermore, we extended our analysis to include advisory or non-binding documents which, while they do not constitute official national AI agendas, are closely related and provide important complementary insights.Footnote 68

Having established a final corpus of national AI agendas, we then identified the sections that are most relevant to the focus of this paper. Given the extensive length and thematic breadth of these documents, which address a wide range of AI-related topics, we adopted a targeted keyword-search methodology that is specifically designed to isolate content that is related to the environment, thereby circumscribing the scope of our analysis. We chose the following keywords to guide our content identification: “environment,” “ecology,” “energy,” “water,” “waste,” “smart cities,” and “data centres.” To account for the diversity of official languages, we took care to translate each such keywords to the official languages in which the national AI agendas of the Member States are written. Where possible, however, we prioritised the use of the English versions of the documents.

Once we had identified the relevant sections through keyword searches, we read these excerpts and analysed them. This step involved contextualising the references to environmental issues with two main ends. First, we try to identify themes and trends that recur across agendas. For example, all of the Member States acknowledge that AI has benefits for environmental protection. Second, we seek to assess the degree of alignment between national agendas and the overarching normative framework of EU law. When we pinpoint occasions on which the Member States depart from the prescriptions that are found in EU policy documents, we point to gaps in policy coherence.Footnote 69 For example, some countries restrict their treatment of the environmental impact of AI to carbon emissions, while others also account for, say, social equity, in their AI strategies.

In our analysis of the national AI agendas, we identified two common motifs. First, the environmental impact of AI is often treated as one of several elements of sustainability. While the ecological ramifications of AI technologies are recognised, they are frequently assessed by reference to a narrow and siloed concept of environmental protection. There is thus a risk that those who make policy and law may fail to appreciate how AI can simultaneously influence and interact with other critical dimensions of sustainability, such as economic development and distributive fairness.Footnote 70 Second, the national agendas usually go beyond carbon emissions.Footnote 71 While the carbon footprint of AI technologies is undoubtedly important,Footnote 72 the environmental implications of said technologies also include resource consumption, waste generation and biodiversity loss. Although only a few countries elaborate on these matters,Footnote 73 among them, there appears to be a general consensus on the broader eco-footprint of AI technology. However, it is important to note that not all national agendas embrace the perspective of AI as a polluting technology; some national agendas simply omit to mention the polluting effects of AI.Footnote 74

Having identified these two main motifs, we categorise Member State attitudes towards AI technology into two distinct main groups. The first group consists of countries that largely overlook the negative environmental impacts of AI. Some of them adopted national AI agendas early and are yet to revise them so as to account for the environmental concerns about AI that have emerged more recently.Footnote 75 Denmark supplies a germane example. While the 2019 National Strategy for Artificial Intelligence does not mention the relationship between AI and environment,Footnote 76 the 2022 National Strategy for Digitalisation acknowledges the need to reduce the climactic and environmental impact of digitalisation.Footnote 77 Other Member States simply choose to focus exclusively on the potential benefits of AI for the environment.Footnote 78 Some of those states have put forward balanced practical recommendations, such as the Austrian call to fund the development of green AI to facilitate the twin transition.Footnote 79 Finally, some Member States channel all of their efforts into non-environmental considerations, such as economic or technological progress.Footnote 80 However, it is noteworthy here to mention the Netherlands, which, while acknowledging the negative environmental impacts of AI and claiming that the benefits significantly outweigh the drawbacks, has largely overlooked the connection between AI and the environment in its national AI agenda.Footnote 81

The countries in the second group recognise the environmental impact of AI and are seeking means of mitigating its adverse impacts proactively.Footnote 82 These nations are obviously much more likely to promote sustainable AI policies. Some of their agendas make explicit references to critical problems such as the rebound effect.Footnote 83 The proposals for solutions which those countries put forward vary widely in content and specificity. Notable proposals are the adoption of stringent environmental standards for data centres,Footnote 84 including the requirement to source (all) their energy from fossil-free sources.Footnote 85 Another proposal is to establish a fund in order to finance the development of energy-efficient AI technology.Footnote 86

Other countries in the second group have opted for a more temperate approach that revolves around assessment and evaluation. The Member States in this subgroup do advocate for cross-national initiatives aimed at developing shared standards and best practices that prioritise environmental considerations and facilitate harmonised deployment across borders, including explicit calls for monitoring the environmental impact of AI technology.Footnote 87 For example, the German National AI Research Agenda calls for the development of a feasible procedure for environmental impact assessments of AI and for the allocation of funding to the systemic collection of empirical data on its CO2-economising potential.Footnote 88 Estonia is alone in focusing chiefly on the role of state rather than that of the private sector,Footnote 89 a tendency which is perhaps attributable to the well-documented successes which the Estonians have achieved by embracing the use of technology in government.Footnote 90 Finally, we should note that there are cases in which the national AI agendas do mention the dual impact of AI but do not articulate any specific programmes to address it.Footnote 91

It is evident from the foregoing that the environmental dimensions of AI are acknowledged in numerous national agendas. However, the highly variegated treatment of the issue evinces the importance of developing a more cohesive approach. Unsurprisingly, this is in line with the recommendations put forth by the European Council in its conclusions of 17 December 2020. There, the European Council indicated that the twin transition can only be completed successfully if the Member States share experiences and lessons from the development and application of AI in the environmental domain.Footnote 92

VI. Conclusion

The regulatory architecture of the EU requires close cooperation between the Union and the Member States. The objectives of the twin transition are such that they demand comprehensive consideration of the environmental impact of AI and the development of a harmonised approach to the resultant problems. The EU recently started laying down a constitutional framework for the twin transition through the EU Declaration on Digital Rights and Principles. The Member States have sought to articulate these soft-law digital sustainability principles through the adoption of national AI agendas. Yet, these policy documents, when viewed in their totality, lack coherence and appear to be at odds on many important issues. For example, some strategies are wholly silent on the negative environmental impacts of AI, while others address these concerns quite comprehensively. On the whole, one cannot avoid the feeling that the Member States, which are supposed to be acting in concert, are providing substantially different solutions to differently defined problems.

Let us conclude by attempting to distil the core causes of this asynchrony. First, although one can observe the first attempts to develop principles of eco-digital constitutionalism at EU level, the Declaration on Digital Rights and Principles remains a political document. Soft law can, of course, be highly influential, but the twin transition appears an unlikely candidate for successful soft-law coordination. AI, a key twin-transition technology, is incredibly contentious. Many countries see it as a vehicle of economic salvation, and there is the idea that national regulatory systems ought to compete with each other for AI investment and R&D. When the problem is reconceptualised in this manner, it is not at all surprising that the national policies which we overviewed are discordant: their content reflects not so much a willingness to give effect to eco-digital constitutional aspirations, such as uniformity of environmental protection, but a desire to secure investment through the competitive provision of growth-oriented national policy.

Second, we observed that several national policies do not account for the negative environmental impact of AI simply because they were written at a time when that impact was known poorly, if at all. This deficiency seems benign because it is fairly easy to address in future iterations of national AI strategies, but it is indicative of a more general and less tractable problem. It is widely recognised that incipient technologies are difficult to regulate because regulators do not know much about their impacts and that technologies whose impact is understood well are difficult to regulate because their unregulated forms are too strongly embedded in markets and societies;Footnote 93 this holds true in the environmental domain as well.Footnote 94 In the case of AI in the EU, this finding is confirmed by the incipient nature of the eco-digital constitutionalisation process and by the fragmentation of regulatory and governance regimes across the Union.

From this perspective, future revisions of national AI policies will benefit from more intensive dialogue between relevant stakeholders who should try to identify better ways to address the interlinked problems which AI development creates for the environment. Relatedly, a more coordinated approach to the twin transition could be highly beneficial; the establishment of the AI Office may play a crucial role in advancing this effort. This having been said, it is important to acknowledge that there is a risk that these topics will be relegated to a secondary role by geopolitics, particularly as the EU attempts to close the industrial gap with technopoles such as the United States or China. The actions that the EU or its Member States perform to shape the development of AI, including the elaboration and operationalisation of digital sustainability principles, can always be perceived as potential obstacles to innovation and industrial development. Constitutionalising eco-digital principles via national AI policies will also imply striking a balance between the preservation of core EU values and the Union’s capability to compete in the fast-moving AI market.

Funding Statement

Open access funding provided by Erasmus University Rotterdam.

References

1 See, e.g., P Dhar, “The Carbon Impact of Artificial Intelligence” (2020) 2 (8) Nature Machine Intelligence 423; E Strubell, A Ganesh and A McCallum, “Energy and Policy Considerations for Deep Learning in NLP” in Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (2019) pp. 3645–50; P Li et al., “Making AI Less Thirsty”(2025) 68 (7) Communications of the ACM 54.

2 See A Perez Victorio and E Celeste, “Eco-Digital Products and Services: Towards New EU Sustainability Rights?” in E Celeste and others (eds), Digital Sovereignty and the Green Transition: EU Challenges in Times of War and Energy Crisis (Hart 2025).

3 I Niet, “EU Public Values Governing the Twin Transition” (2024) 33 (3) RECIEL 452.

4 AS Epstein, “EU Environmental Law in the Digital Age: A Critical Outlook on the Twin Transition’s Legal Structure” (2025) European Journal of Risk Regulation 1.

5 A Quintavalla and O Yalnazov, “EU Law in the Age of the Twin Transition: Risk, Uncertainty, and Legal Form” (2025) 45 (4) Environment Systems and Decisions 65.

6 U Pagallo, JC Sciolla and M Durante, “The Environmental Challenges of AI in EU Law: Lessons Learned from the Artificial Intelligence Act (AIA) with Its Drawbacks” (2022) 16 (3) Transforming Government: People, Process and Policy 359; P. Hacker, “Sustainable AI Regulation” (2024) 61 Common Market Law Review 345.

7 T Le Goff, “Environmental Law’s Principles Applied to Artificial Intelligence: A Path towards Regulation” (2025) International Review of Law, Computers & Technology 1.

8 T Santarius et al, “Digitalization and Sustainability: A Call for a Digital Green Deal” (2023) 147 Environmental Science & Policy 11.

9 See, e.g., E Daly, “Environmental Constitutionalism in Defence of Nature” (2018) 53 Wake Forest Law Review 667.

10 A Polk, “The Unfulfilled Promise of Environmental Constitutionalism” (2022) 74 Hastings Law Journal 123.

11 JR May, E Daly, LJ Kotze and C Soyapi, New Frontiers in Environmental Constitutionalism (United Nations Environment Programme 2017).

12 Robinson Township v Commonwealth, 83 A.3d 901 (Pa 2013).

13 On the Robinson Township case, see E Daly and JR May, “Robinson Township v Pennsylvania: A Model for Environmental Constitutionalism Symposium: Global Environmental Constitutionalism” (2015) 21 Widener Law Review 151.

14 See, e.g., FL Maldonado and K Yánez, “El constitucionalismo ambiental en Ecuador” (2020) 97 Actualidad Jurídica Ambiental 5.

15 See DR Boyd, “Catalyst for Change: Evaluating Forty Years of Experience in Implementing the Right to a Healthy Environment” in JH Knox and R Pejan (eds), The Human Right to a Healthy Environment (Cambridge University Press 2018).

16 Please note that the Maastricht Treaty of 1992 strengthened the commitment of the Community in virtue of the express reference to “sustainable and non-inflationary growth respecting the environment” (Article 2).

17 On regional environmental constitutionalism, see LJ Kotzé, “The Transnationalization of Environmental Constitutionalism” in V Heyvaert and LA Duvic-Paoli (eds), Research Handbook on Transnational Environmental Law (Edward Elgar 2020) at 173–5.

18 Ibid; CT Reid, “The Court of Justice and Protected Areas in the European Union” in G Antonelli et al. (eds), Environmental Law Before the Courts: A US–EU Narrative (Springer 2023).

19 D Amirante, “Environmental Constitutionalism through the Lens of Comparative Law: New Perspectives for the Anthropocene” in D Amirante and S Bagni (eds), Environmental Constitutionalism in the Anthropocene (Routledge 2022) 152.

20 While the Greek constitution considered environmental protection as an obligation of the State – which also constitutes a collective right enjoyed by everyone, Art. 66 of the Portuguese constitution enshrined the “right to a human, healthy and ecologically balanced environment” and the “duty to defend it,” and Art. 45 of the Spanish constitution established the “right to enjoy an adequate environment for the development of the person, as well as the duty to maintain it”. See ibid 155.

21 Ibid 157–63.

22 On this reform, see D Greco, “The 2022 ‘Environmental Reform’ of the Italian Constitution and International Law” (2023) 32 The Italian Yearbook of International Law Online 263.

23 The Charter is a comprehensive text consisting of ten articles that define the rights, principles and duties related to the environment, and thus reflects a strengthened commitment to systematically enshrine environmental protection within the constitutional framework. See R O’Gorman, “Environmental Constitutionalism: A Comparative Study” (2017) 6 Transnational Environmental Law 435, at 450. Amirante (n 19) at 162.

24 See, e.g., E Celeste, “Conceptual Approaches to Digital Constitutionalism: A Counter-Critique” in I Spiecker gen. Döhmann, L Schertel Mendes and RR Campos (eds), Digital Constitutionalism (Nomos 2025) 15–46.

25 A Quintavalla and J Temperman, Artificial Intelligence and Human Rights (Oxford University Press 2023).

26 For a comprehensive reconstruction, see Celeste (n 24).

27 See O Pollicino, Judicial Protection of Fundamental Rights on the Internet: A Road towards Digital Constitutionalism? (Hart 2021); G de Gregorio, Digital Constitutionalism in Europe: Reframing Rights and Powers in the Algorithmic Society (Cambridge University Press 2022).

28 Ibid; see also, E Celeste and G Formici, “Constitutionalizing Mass Surveillance in the EU: Civil Society Demands, Judicial Activism, and Legislative Inertia” (2024) 25 German Law Journal 427.

29 V Papakonstantinou and P De Hert, “The Vagelis Papakonstantinou and Paul De Hert, ‘The Regulation of Digital Technologies in the EU: The Law-Making Phenomena of ‘Act-Ification’, ‘GDPR Mimesis’ and ‘EU Law Brutality’” (2022) Technology and Regulation 48.

30 Ibid.

31 L Floridi, “The European Legislation on AI: A Brief Analysis of Its Philosophical Approach” (2021) 34 Philosophy & Technology 215.

32 European Commission, Communication “Establishing a European Declaration on Digital rights and principles for the Digital Decade” COM/2022/27 final.

33 For a detailed analysis of this concept see Celeste and Perez Victorio (n 2).

34 Speech in the European Parliament Plenary Session Ursula von der Leyen President-elect of the European Commission, Strasbourg, 27 November 2019. Available at <https://www.statewatch.org/media/documents/news/2019/nov/eu-com-von-der-leyen-ep-vote-speech-27-11-19.pdf>.

35 European Commission, Communication The European Green Deal, COM(2019) 640 final, at 7; see also European Commission, Communication Shaping Europe’s digital future, COM(2020) 67 final, at 1.

36 European Commission, Communication “A New Industrial Strategy for Europe,” COM(2020) 102 final and European Commission, Communication “Updating the 2020 new industrial strategy: Building a stronger single market” COM/2021/350 final.

37 European Council, Draft Council conclusions on digitalisation for the benefits of the environment, 13957/20.

38 Epstein (n 4) at 1ff.

39 P Gailhofer et al., The role of Artificial Intelligence in the European Green Deal, Study for the special committee on Artificial Intelligence in a Digital Age (Policy Department for Economic, Scientific and Quality of Life Policies, European Parliament 2021) available at <https://www.europarl.europa.eu/RegData/etudes/STUD/2021/662906/IPOL_STU(2021)662906_EN.pdf>.

40 European Commission, Communication “Towards a green, digital and resilient economy: our European Growth Model” COM(2022) 83 final at 6; European Commission, Communication “2022 Strategic Foresight Report. Twinning the green and digital transitions in the new geopolitical context,” COM(2022) 289 final at 2ff.

41 EU Declaration on Digital Rights and Principles, Recital 6. For the “external” role of the Declaration, see Recital 11.

42 On the EU Declaration, see E Celeste, “Digital Constitutionalism, EU Digital Sovereignty Ambitions and the Role of the European Declaration on Digital Rights” in A Engel, X Groussot and G Thor Petursson (eds), New Directions in Digitalisation: Perspectives from EU Competition Law and the Charter of Fundamental Rights (Springer 2025).

43 For a reconstruction of these three principles, see E Celeste and A Perez Victorio, “Eco-Digital Constitutionalism? Analyzing the Emergence of the Principles of Sustainable Digitalization” ICON•S 2025 Annual Conference. See also Perez Victorio and Celeste (n 2).

44 See, e.g., European Commission (n 32) and European Commission (n 36).

45 Pagallo et al (n 6); Hacker (n 6); JS Carbonell, In Search of the Twin Transition: The Limited Performativity of the “Green and Digital” Transitions in the European Automotive Industry (JRC Working Papers Series on Labour, Education and Technology 2024); E Celeste and G Dominioni, “Digital and Green: Reconciling the EU Twin Transitions in Times of War and Energy Crisis” in F Fabbrini and CA Petit (eds), Research Handbook on Post-Pandemic EU Economic Governance (Edward Elgar 2024).

46 A Perez Victorio, E Celeste and A Quintavalla, “Greening AI? The New Principle of Sustainable Digital Products and Services in the EU” (2024) 61 (4) Common Market Law Review 1019.

47 Hacker (n 5).

48 Art. 7, 8, 12, 28 and 31 of the EU AI Act. N Alder, K Ebert, R Herbrich and P Hacker, “AI, Climate, and Transparency: Operationalizing and Improving the AI Act,” 28 August 2024, at (accessed 10 September 2024); T. Le Goff, “AI-driven Sustainability in the Energy Sector: Regulatory and Governance Issues” (2025) Oxford Intersections: AI in Society.

49 Directive (EU) 2023/1791 of the European Parliament and of the Council of 13 September 2023 on energy efficiency and amending Regulation (EU) 2023/955, OJ L 231, 20.9.2023, p. 1, Art. 12 and Annex VII.

50 Perez Victorio et al (n 46).

51 F Fabbrini, “The Principle of Subsidiarity” in R Schütze and T Tridimas (eds), Oxford Principles of European Union Law: The European Union Legal Order: Volume I (Oxford University Press 2018) 221–42.

52 See, e.g., A Quintavalla, “Artificial Intelligence and the Right to a Healthy Environment” in A Quintavalla and J Temperman (eds), Artificial Intelligence & Human Rights (Oxford University Press 2023) 425–38; RB Sander, “Confronting Risks at the Intersection of Climate Change and Artificial Intelligence: The Promise and Perils of Rights-Based Approaches” (2025) Netherlands Quarterly of Human Rights, 09240519251336962.

53 AI Act, Art. 95.

54 Similarly, one should point at the significant discretion left to Member States in terms of enforcement and supervision.

55 European Commission, Communication “Coordinated Plan on Artificial Intelligence” COM/2018/795 final.

56 Ibid, at 9.

57 Ibid.

58 European Commission, Communication “Fostering a European approach to Artificial Intelligence” COM/2021/205 final.

59 Ibid., Annex, at 37–8.

60 Ibid., Annex, at 26 and 31.

61 Ibid., Annex, chapter 11.

62 Ibid., Annex, at 40.

63 Ministerial Declaration, “A Green and Digital Transformation of the EU” Digital Day 2021, March 19, Porto (Portugal). Available at <https://digital-strategy.ec.europa.eu/en/news/eu-countries-commit-leading-green-digital-transformation>.

64 Toulouse Call for a Green and Digital Transition in the EU. Available at <https://www.economie.gouv.fr/files/files/2022/Call_for_Green_Digital_Transition_EU.PDF>.

65 European Commission, Communication “Coordinated Plan on Artificial Intelligence” COM/2018/795 final; European Commission, Communication “Fostering a European approach to Artificial Intelligence” COM/2021/205 final.

66 R Jorge Ricart, V Van Roy, F Rossetti and L Tangi, AI Watch. National Strategies on Artificial Intelligence: A European Perspective (Publications Office of the European Union 2022).

67 See infra.

68 Ibid.

69 X Gao, “The EU’s Twin Transitions Towards Sustainability and Digital Leadership: A Coherent or Fragmented Policy Field?” (2025) 59 (1) Regional Studies 2360053.

70 As an exception, see Germany’s National AI Research Agenda at 9; Spain’s National AI Research Agenda at 20 and 55; and Slovenian Government, Digital Slovenia 2030 Strategy (2023) <https://nio.gov.si/en/products/strategija%2Bdigitalna%2Bslovenija%2B20300?release=0.1> at 10.

71 A Faggian, A Marzucchi and S Montresor, “Regions Facing the ‘Twin Transition’: Combining Regional Green and Digital Innovations” (2025) 59 (1) Regional Studies 2398555 at 2.

72 See, e.g., Irish Government, Ireland’s National AI Strategy: AI – Here for Good Refresh 2024 (2024) at 30–1 <https://enterprise.gov.ie/en/publications/publication-files/national-ai-strategy-refresh-2024.pdf>.

73 Swedish Government, Sveriges digitaliseringsstrategi 2025–2030 (2025) at 54 <https://www.regeringen.se/contentassets/fe3e296228fb474f803a986ae3842b4c/sveriges-digitaliseringsstrategi-20252030.pdf>; Croatian Government, Digital Croatia Strategy for the period until 2032 <https://mpudt.gov.hr/UserDocsImages/RDD/SDURDDdokumenti/Strategija_Digitalne_Hrvatske_final_v1_EN.pdf?lang=de> at 11 and 99; Greek High-level Advisory Committee on Artificial Intelligence, A Blueprint for Greece’s AI Transformation (2024) at 147 <https://foresight.gov.gr/wpcontent/uploads/2024/11/Blueprint_GREECES_AI_TRANSFORMATION.pdf>.

74 See below.

75 AI Portugal 2030; Policy for the Development of Artificial Intelligence in Poland from 2020 at 8 and 33; Concept for the Development of Artificial Intelligence in Bulgaria until 2030 at 13-15; Lithuanian Government, Artificial Intelligence Strategy: A Vision of the Future (2019) <https://eimin.lrv.lt/uploads/eimin/documents/files/DI_strategija_ENG(1).pdf>; Latvian government, Par mākslīgā intelekta risinājumu attīstību (2020) <https://likumi.lv/ta/id/342405-par-maksliga-intelekta-risinajumu-attistibu>; Slovakian Government, 2030 Digital Transformation Strategy for Slovakia (2019) <https://mirri.gov.sk/wp-content/uploads/2019/10/SDT-English-Version-FINAL.pdf>; Cyprus Government, EθNική ΣTραTηγική TEχNηTής NoημoσύNης (TN): ΔράσEις για TηN αξιoπoίηση και αNάπTυξη Tης TN σTηN κύπρo (2020) https://www.gov.cy/app/uploads/sites/13/2024/04/EθNική-ΣTραTηγική-TN.pdf>.

76 For exceptions, see Danish Government, National Strategy for Artificial Intelligence (2019) at 33 and 44 <https://en.digst.dk/media/lz0fxbt4/305755_gb_version_final-a.pdf>.

77 Danish Government, National Strategy for Digitalisation: Together in the digital development (2022) at 39–40 <https://en.digst.dk/media/mndfou2j/national-strategy-for-digitalisation-together-in-the-digital-development.pdf>.

78 Artificial Intelligence Strategy of the Austrian Federal Government, Annex at 5–6; Romanian Government, Strategia națională în domeniul inteligenței artificiale 2024–2027 (2024) at 85 <https://www.mcid.gov.ro/wp-content/uploads/2024/03/SN-IA-final.pdf>; Hungarian Government, Magyarország Mesterséges Intelligencia Stratégiája (2025–2030) (2025) <https://cdn.kormany.hu/uploads/document/c/c0/c0d/c0dfdbd37cfa520ae37361a168d244c85e7295af.pdf> at 80; 90–1.

79 Artificial Intelligence Strategy of the Austrian Federal Government at 42.

80 Dutch government, The Netherlands’ Digitalisation Strategy Joint acceleration is the only option (2025) <https://www.nldigitalgovernment.nl/wp-content/uploads/sites/11/2025/09/108.201-NDS-publicatie_Engels_v6-WEB-1.pdf>; Finnish government, Suomesta voittaja kaksoisiirtymässä – tavoitteista käytäntöön: Tekoäly 4.0 -ohjelma, toinen väliraportti (2021) at 19–20 <https://julkaisut.valtioneuvosto.fi/items/40c3013b-1caa-4544-a0c5-bd47e20e2d99>.

81 Ministerie van Binnenlandse Zaken en Koninkrijksrelaties, Letter to the Parliament “Government-wide vision on generative AI of the Netherlands” (2024) at 11 <https://www.government.nl/documents/parliamentary-documents/2024/01/17/government-wide-vision-on-generative-ai-of-the-netherlands>.

82 See, e.g., Artificial Intelligence Strategy of the German Federal Government 2020 Update; France 2030: National Strategy for Artificial Intelligence; Spain’s National AI Research Agenda; Belgian government, National convergence plan for the development of artificial intelligence: Becoming an AI Smart Nation (2022) at 26–7 <https://bosa.belgium.be/sites/default/files/content/images/DigitaleOverheid/AI/Plan_AI_EN.pdf>; The Government of the Grand Duchy of Luxembourg, Luxembourg’s AI Strategy (2025) at 19 <https://gouvernement.lu/fr/publications/rapport-etude-analyse/minist-digitalisation/2025-luxembourg-ai-strategy-en.html>.

83 Artificial Intelligence Strategy of the German Federal Government 2020 Update at 15–6; France’s Villani Report.

84 The Government of the Grand Duchy of Luxembourg, Luxembourg’s Data Strategy (2025) at 32 <https://gouvernement.lu/fr/publications/rapport-etude-analyse/minist-digitalisation/2025-luxembourg-data-strategy-en.html>.

85 Swedish Government, AI-kommissionens Färdplan för Sverige (2025) at 24 <https://www.regeringen.se/rapporter/2024/11/ai-kommissionens-fardplan-for-sverige/>; Irish Government (n 72); Croatian Government (n 73).

86 France 2030: National Strategy for Artificial Intelligence at 26.

87 Swedish Government (n 85) at 24; Artificial Intelligence Strategy of the German Federal Government 2020 Update at 15–6; Slovenian Government (n 70) at 51; Maltese Digital Innovation Authority, Malta Towards Trustworthy AI: Malta’s Ethical AI Framework (2019) at 22 <https://mdia.gov.mt/national-strategies/malta-ai-strategy-and-vision/>.

88 Artificial Intelligence Strategy of the German Federal Government 2020 Update at 15–6.

89 Estonian Government, Estonia’s Digital Agenda 2030 (2021) at 19 <https://www.mkm.ee/sites/default/files/documents/202204/Digiühiskonna%20arengukava_ENG.pdf>.

90 Ibid.

91 Spain’s National Artificial Intelligence Strategy; Belgian government (n 82).

92 The Environment Council of 17 December 2020 addressed the twin societal challenge of digital transformation and green transition in Digitalisation for the Benefit of the Environment – Council conclusions 17 December 2020. Please also note that, in March 2021, twenty-four Member States, Norway and Iceland signed a declaration to accelerate the use of green digital technologies for the benefit of the environment, i.e., by encouraging the development and use of energy-efficient algorithms.

93 D Collingridge, The Social Control of Technology (Frances Pinter 1980); R Leenes, “Of Horses and Other Animals” (2019) Technology & Regulation 1; MK Kołacz, A Quintavalla and O Yalnazov, “Who Should Regulate Disruptive Technology?” (2019) 10 (1) European Journal of Risk Regulation 4.

94 A Quintavalla and O Yalnazov, “Regulating Eco-Innovation in the European Union” (2025) 68 (3) Journal of Environmental Planning and Management 539.