The Moderating Role of Governmental Artificial Intelligence in Shaping Green Growth Dynamics in the European Union

Abstract

In the contemporary era of accelerated artificial intelligence (AI) development, governments are progressively embedding AI technologies within administrative structures and public service delivery systems, thereby influencing green growth through a variety of direct and indirect mechanisms. This study examines the impact of governmental artificial intelligence, captured by readiness of governments for the adoption of artificial intelligence, alongside key macroeconomic, social, and institutional governance factors, on green growth across European Union countries between 2019 and 2023. Employing a combination of static and dynamic panel econometric techniques (OLS, FGLS, PCSE, LIML-IV, and system GMM), the analysis reveals that governmental deployment of AI exerts a positive effect on green growth within the sample. The empirical results demonstrate that governments’ readiness to systematically deploy AI exerts a positive influence on green growth, thereby supporting the hypothesis that governmental AI capabilities foster sustainable development. The positive coefficient on the lagged green growth, together with the stable significance levels across all model specifications, provides quantitative evidence that governmental AI readiness has a positive and statistically significant effect on green growth. The synergistic interaction among macroeconomic, social, and institutional governance factors provides strong empirical support for the design of robust policies that promote the implementation of AI at the governmental level, coupled with mechanisms to stimulate economic expansion, public investment, and foreign capital inflows.

1. Introduction

All facets of human life, business, or the governmental sphere are being significantly impacted by the recent exponential growth of artificial intelligence (AI). Over the coming decade, AI is expected to contribute to an average annual increase of around 7% in global GDP, largely attributable to efficiency gains and productivity enhancements in the labor force [1]. AI divides the scholarly community between those who argue that it is driving a paradigm shift in growth models by replacing labor—a limited production factor—with AI-generated resources that operate as unlimited and infinitely scalable inputs [2,3,4] and those who emphasize that, for growth to be sustainable given the still-immature state of technology, AI must complement human labor by enhancing efficiency, improving quality, and enabling the performance of novel tasks [5]. Some adverse effects of AI on economic growth may arise from technological limitations, legal and regulatory constraints, and socio-professional disruption, as well as organizational and institutional rigidities [6].

In spite of potential economic positive effects, AI development raises questions about its relationship with the environment and sustainable development. Consequently, a strand of research has emerged focusing on sustainable AI, encompassing both the sustainability of AI itself and the role of AI in promoting sustainable development [7]. The sustainability of AI can be assessed through life-cycle analysis, which distinguishes between the software and hardware components. In recent years, increasing attention has been devoted to the environmental implications of AI development, including rising energy, water, and mineral consumption, higher carbon emissions, and larger volumes of waste [7,8,9,10,11,12,13,14]. At the same time, the use of AI in support of sustainable development remains an area of active debate, where scholars assess its capacity to contribute to the achievement of the Sustainable Development Goals [15,16,17,18,19,20]. Recent empirical evidence also suggests that investments in AI can positively support sustainable development [21].

The orientation of research toward integrating the economic, social, and environmental impacts of AI has generated new conceptual frameworks gravitating around green growth, green economy, and green total factor productivity [22,23]. While AI creates substantial opportunities for advancing the green economy, it also raises important questions, such as the high energy consumption of AI systems, data privacy worries, and unintentional environmental costs of AI adoption [24]. Developing the responsible and ethical use of AI is therefore crucial to fully exploit its capacity in promoting sustainability and fostering a greener economy, given that its environmental externalities cannot be ignored [24].

In this context, governments face a dual challenge. On the one hand, they must design and implement strategies and guidelines to ensure responsible AI that adheres to ethical and societal principles [25,26]. On the other hand, governments are tasked with integrating AI into digital public services [27]. Such integration can yield significant benefits, including reducing corruption, improving public finance management, enhancing the quality of policy decisions, and increasing both the quality and accessibility of public services [28]. The adoption of AI by governments has the potential to modernize public administration and contribute to improving citizens’ well-being. However, it also entails risks related to personal data security, algorithmic bias, potential misuse, and perceptions of illegitimacy [27].

Despite the growing relevance of artificial intelligence in public sector governance, however, there is a significant gap in the literature on its impact on green growth. Understanding how AI adoption in governments influences green growth is an underexplored topic. Different approaches to AI adoption in government can be observed globally. For instance, European Union governments follow strategies that differ from those of the United States in promoting AI integration into public policies [29]. Consequently, a focused study on European Union countries is necessary to assess the impact of governmental AI initiatives on green growth.

Based on these considerations, this study aims to examine how governments’ readiness to implement artificial intelligence (AI) influences green growth in European Union countries. The main objective is to analyze the structural role of government AI readiness in shaping environmentally sustainable growth trajectories, emphasizing how the preparedness of public institutions to adopt and integrate AI can act as a critical determinant of green growth. Accordingly, the research seeks to address the following research question: How does governments’ readiness for artificial intelligence (AI) adoption affect green growth? To answer this question, this study investigates both the static and dynamic effects of the governments’ AI readiness on green growth by employing a set of econometric techniques, using the Government Artificial Intelligence Readiness Index as the core explanatory variable [30] and Green Growth Index as the dependent variable [31]. In doing so, the analysis incorporates a broad set of determinants that may condition or mediate this relationship, including macroeconomic factors (gross domestic product, general government final consumption expenditure, and foreign direct investment net inflows), social factors (the Human Development Index and degree of urbanization) and institutional factors (the Political Stability and Violence/Terrorism Absence Indicator). This comprehensive framework captures the multifaced channels through which AI governments’ readiness can affect green growth, providing novel insights between digital governance and environmental sustainability.

This research adds several contributions to the literature. First, the interdisciplinary nature of the research, at the intersection of digital technologies, public governance, and green economics, makes it both timely and impactful, offering not only theoretical advancements but also practical guidance for national and international sustainability strategies. Existing studies on AI and sustainability largely focus on the business sector, without highlighting the role of public institutions in shaping green growth. This research emphasizes the major role of the public sector in sustainable development. Moreover, the insights gained from such research can contribute to a broader debate on global climate governance, in which governments play a central role in coordinating collective actions and meeting international commitments. Second, this study employs a multidimensional framework of macroeconomic, social, and institutional factors, which can moderate and modulate the channels through which AI governments’ readiness influences green growth. Third, the study analyzes a special conglomerate of states, represented by the countries of the European Union, that exhibit structural heterogeneity within a common European framework and that develop differently economically, socially, and administratively. These countries, while retaining sovereignty over internal public and governmental policies, are obliged to harmonize those policies with Community rules and thus acquire the preconditions for a high degree of governmental homogeneity. Fourth, the research leverages one of the few longitudinally consistent indicators that captured governments’ preparedness for AI adoption, represented by the Government Artificial Intelligence Readiness Index, developed by Oxford Insights [30]. In this way, it contributes uniquely to the literature, which is otherwise constrained by the scarcity and recent development of comparable measures on AI adoption in governments.

Next, the study presents the theoretical background regarding the influence of governmental artificial intelligence on green growth, as well as the influence exerted by set of macroeconomic, social, and institutional governance factors. Section 3 contains the data, sample considerations, research methodology, and models. The results are presented in Section 4, and discussions on the interactions between the analyzed variables are presented in Section 5. The final part of the study covers the conclusions, policy implications, limitations, and further directions.

2. Literature Review

2.1. The Relationship Between Readiness of Governments for the Adoption of Artificial Intelligence and Green Growth

The rapid evolution of AI has established it as a profoundly transformative technology, requiring strategic adaptation across all sectors, particularly within public administrations worldwide [32,33]. The effective integration and utilization of AI by governmental bodies is encapsulated by the concept of “Government AI Readiness”, which has become a critical focus of both academic research and practical policy debate [34,35]. The Government AI Readiness Index, developed by Oxford University through the Oxford Insights initiative, represents a multi-dimensional framework for assessing a country’s preparedness to address both the opportunities and challenges posed by AI [30,35,36]. Conceptually, the index is typically structured around three main pillars and 40 indicators, which capture the multi-dimensional nature of AI readiness and enable comparative evaluation across countries. The government pillar evaluates the strategic vision, regulatory and ethical frameworks, and the digital capacities required for AI development and governance. The technology pillar encompasses the maturity of AI tools, innovation capacity, and the availability of human capital with relevant AI skills. The data and infrastructure pillar assesses access to high-quality data and the foundational infrastructure necessary for AI adoption [35,36].

AI in the public sector entails far more than the development of digital tools. Governments must adopt a multi-criteria approach that simultaneously addresses strategic priorities, ethical considerations, technological requirements, infrastructural capacity, and data governance when designing an AI framework. A government’s approach to AI strategies is a core determinant of its readiness to implement AI in public services. The formulation of national and international AI strategies often generates heated debate among civil community, policymakers, and political leaders. While many risks of AI implementation—such as threats to privacy, algorithmic bias and potential misuse—are well-documented, the public value must prevail over narrow or sectoral interests. Accordingly, government policies in the AI domain should prioritize the creation of clear regulatory and ethical outlines that guarantee responsibility, equity, and public confidence.

Successful implementation of AI at the government level requires not only technical expertise but also administrative capacity. The limited accessibility of cutting-edge AI technologies, their high costs, and varying levels of institutional capacity across countries create substantial heterogeneity in governments’ ability to adopt and deploy AI systems. The development of both the soft and hard components of governmental AI typically depends on state procurements from domestic and international developers and suppliers, as well on public investment in research and development conducted in collaboration with industry and universities. However, without adequate human capital—specialists trained to design, implement and manage AI solutions—the process remains highly challenging. Moreover, the successful provision of AI enables public services presupposes that citizens themselves possess a minimum level of digital literacy and acceptance, which underscores the role of education and training as critical dimension of the governmental AI architecture.

Once government-specific AI technologies are developed, infrastructural support becomes equally essential. Substantial public investment is required for the acquisition of equipment capable of sustaining AI systems across multiple fields of application, alongside the provision of robust digital infrastructure, including broadband networks, cloud computing resources, and cybersecurity frameworks. Finally, the datasets generated through the use of governmental AI raise complex challenges concerning access, retention, security, and deletion, each of which demand rigorous governance mechanisms.

Research into AI readiness and capabilities in public organizations draws on several theoretical frameworks that help explain the complex set of factors involved. The Technology–Organization–Environment (TOE) framework is one of the most widely applied models for understanding the adoption and diffusion of technological innovations [32,37]. Its relevance for analyzing AI diffusion in public organizations derives from its ability to integrate both internal and external contextual aspects that shape technology assimilation [32,38]. The TOE framework identifies three key contexts: technological context (referring to the characteristics of AI technologies themselves, including their perceived benefits and complexities), organizational context (capturing internal factors such as organizational size, structure, managerial support, and innovativeness), and environmental context (including external pressures such as governmental policies, regulatory frameworks, industry competition, and citizen expectations) [32]. The flexibility of the TOE framework allows researchers to adapt and extend it to incorporate specific variables relevant to the type of AI technology or to the characteristics of public sector organizations [39,40]. Some authors applied such an extended model to investigate factors influencing the development of AI capabilities in European municipalities [32].

The conceptualization of AI readiness is also closely linked to the Resource-Based View (RBV) of the firm [41]. RBV argues that an organization’s sustainable competitive advantage derives from its unique and valuable bundle of resources [42]. In the case of AI, this means that acquiring technology alone is insufficient; public organizations must also strategically cultivate and integrate complementary resources—tangible, human, and intangible—to successfully deploy AI systems and generate value [32,34]. This framework highlights the need for a holistic approach to AI readiness, going beyond superficial adoption and ensuring deep organizational integration.

Green growth, defined as economic development that decouples from environmental degradation through resource efficiency, innovation, and clean technologies, represents a pattern change from conventional growth models [43,44]. The Green Growth Index, developed by the Global Green Growth Institute [31] represents a fundamental paradigm shift in development, emerging from the recognition that traditional economic indicators, such as the Gross Domestic Product (GDP), are inadequate for assessing sustainable progress in the context of environmental degradation, resource depletion, and climate change [45,46]. We argue that the Green Growth Index provides a more comprehensive framework by linking economic performance with ecological and social dimensions, thereby filling the gaps left by GDP. The Green Growth Index constitutes a major advance over conventional economic indicators by embedding sustainability into the evaluation of development. We argue that its importance lies not only in its conceptual design but also in its potential to reshape policy priorities by making sustainability measurable and comparable across contexts. This composite index captures countries’ performance in terms of sustainable development, by linking to global targets such as the Sustainable Development Goals, Paris Climate Agreement, and Aichi Biodiversity Goals [31]. The index captures around 40 indicators grouped in four dimensions: efficient and sustainable resource use, natural capital protection, green economic opportunities, and social inclusion [31].

Several theoretical perspectives inform the conceptualization of the links between technological advanced and green growth. Dynamic Capabilities Theory (DCT) stresses that long-term organizational performance depends on the ability to adapt, reconfigure, and integrate existing resources in dynamic environments [47,48,49,50]. In the context of green growth, technologies are conceptualized as dynamic capabilities that can enhance performance while promoting sustainability. We consider this interpretation particularly valuable, as it highlights how innovation enables the transition away from unsustainable practices [50,51]. The Business Technology Adoption Model explains the uptake of innovations through perceived benefits, organizational readiness, and external pressures [50,52]. We concur with this view, as it reinforces the idea that institutional and structural conditions are decisive for whether firms embrace green innovation. The Resource Curse Theory, by contrast, highlights the paradox whereby resource-rich countries in the EU often face slower growth, inequality, and vulnerability to shocks, rather than prosperity [53]. In this regard, we maintain that green growth functions as an antidote by encouraging diversification, renewable energy investments, and sustainable practices [53]. The Structural Dividend Hypothesis posits that shifting from heavy industries to high-tech eco-friendly sectors enhances productivity while lowering emissions [54]. Finally, research on carbon taxation shows that while such measures reduce emissions and provide socio-economic benefits, achieving the so-called “triple dividend” of growth, equity, and sustainability simultaneously is rare. We note that more commonly “double dividends” emerge, which highlights the inherent trade-offs in green growth strategies [55].

Based on these considerations, we can configure the research hypothesis for this study:

Hypothesis 1 (H₁).

The readiness of governments for the adoption of artificial intelligence positively influences green growth.

2.2. The Role of Economic, Social, and Governance Factors in Mediating the Effects of Governmental Artificial Intelligence on Green Growth

AI in government promotes green growth not only directly but also indirectly through its effects on economic efficiency, social inclusiveness, and governance quality, which act as critical mediating channels. This section describes main macroeconomic, social, and governance factors that modulate the influence of governmental AI on green growth: economic growth, general government final consumption expenditure, foreign direct investment, human development, urbanization, and political stability.

By integrating AI, the government aims to turn economic growth into an engine of green growth, through mechanisms for efficient resource allocation, stimulation of eco-innovation, optimization of fiscal policies, strengthening human capital, and strengthening institutional transparency and governance. The debate on sustainable development increasingly emphasizes the complex relationship between economic growth—traditionally measured by GDP—and the imperative for environmentally responsible practices encapsulated in the notion of green growth. The literature reveals a complex and multifaceted relationship between economic growth and green growth. Green growth policies—when supported by sound governance, innovation, and robust environmental regulation—can mitigate degradation and foster sustainable development, particularly in resource-dependent economies. Nonetheless, economic expansion can intensify vulnerabilities to the resource curse and exacerbate environmental pressures in early development stages.

It is debated whether sustained economic growth, in particular GDP expansion, can be reconciled with green growth. The mechanisms that might explain such an interaction are typically attributed to high-income countries that implement cleaner technologies and are described through the concept of absolute decoupling, a situation in which the percentage increase in GDP is outweighed by a larger percentage reduction in the intensity of CO₂ emissions per unit of GDP [56]. On the other hand, in emerging countries, such as those in Eastern Europe, economic growth is harmful to the environment, given the contribution of economic activity in general to increased emissions and amplified energy consumption, which, if not supported by renewable sources, degrades the environment and produces pollution [57]. Absolute decoupling is generally assumed to rely on the capacity generated by GDP growth to finance and implement cleaner technologies and infrastructures, enabling the replacement of existing systems with lower-carbon or more energy-efficient alternatives [56]. The drivers of absolute decoupling are commonly associated with structural and technological transformations, particularly the expansion of renewable energy, the tertiarization of economies, and the implementation of ambitious and coherent environmental policies [58]. A higher GDP per capita can, paradoxically, increase vulnerability to the resource curse through demand for resource-intensive goods [53]. We highlight this paradox as a reminder that growth-led sustainability is neither linear nor automatic. Carbon emissions directly undermine green growth [50], although evidence also shows that green growth contributes to reducing environmental degradation [59].

The implementation of AI in governments enhances the capacity to prioritize environmentally sustainable investments, improve supply chain transparency, and align public expenditure with green growth objectives. High-quality government efficiency and robust governance institutions significantly foster green economic growth [60,61]. Government final consumption expenditure can significantly positively impact green growth by several mechanisms, such as financing green investments and public goods, promoting sustainable economic cycles via green oriented policies and regulations, and contributing to human capital development, education, and governmental institutional efficiency as catalysts for green growth and sustainability. Public expenditure plays a complex and crucial role in affecting and stimulating green economic growth, in particular through its composition and targeting [62]. Although the topic is of undeniable importance, the relationship between general government final consumption expenditure and green growth remains largely undiscovered, with the existing literature offering only scarce and fragmented contributions. Previous studies have shown that the topic still requires considerable evidence and detailed analysis [62,63].

Different categories of public expenditure can generate mixed effects on green growth, as some spending items foster sustainability and innovation, while others may reinforce environmentally harmful patterns. Government public spending can influence green growth by channeling resources into sustainable sectors, which stimulates green industries, fosters innovation, and generates employment. Government final consumption expenditures may significantly affect green economic growth, as increased spending on renewable energy, sustainable infrastructure, and green technology research and development can boost related industries, create jobs, and support the expansion of the green economy [64].

The specific channels through which public spending on research and development and education positively impact green growth appear to operate primarily via technological innovation and the accumulation of human capital [46,65]. Public spending on education, research, and development is perceived as having a positive impact on sustainability and the growth of the green economy [62]. Lower social welfare expenditures, such as those in non-EU countries, can help reduce production costs and thus promote the green economy [64]. On the other hand, public expenditures in information technology may harm the environment and hinder green growth, potentially intensifying environmental degradation, by stimulating energy intensive production, accelerating electronic waste generation, and increasing the overall carbon footprint of digital infrastructures [61].

Overall, the impact of government final consumption expenditure on green growth is complex and context-dependent, with positive effects emerging when resources are directed toward green investments, education, and innovation, but potential negative outcomes arising from spending that reinforces energy-intensive or environmentally harmful activities. Given the scarcity and fragmentation of the existing research, further empirical investigation is essential to disentangle these heterogenous effects and clarify the specific mechanisms through which public spending shapes green growth.

The relationship between foreign direct investment and green economic growth has attracted significant scholarly attention, being the subject of numerous empirical and theoretical studies [66,67]. The impact of foreign direct investment on green growth is neither uniformly positive nor negative, but there are various factors that nuance the interaction, especially depending on the characteristics of the host country, its governance, the nature of the investments, etc. [66,67]. Foreign direct investment could act as driver for promoting inclusive green growth [66,67]. Multinational enterprises engage in the transfer of technology, humans, and infrastructure to host economies, which may, via spillover effects, promote the efficient allocation of resources and support the transition towards green growth [68,69,70,71].

On the other hand, recent research argues that foreign direct investment hinders green economic growth [72,73,74]. Foreign direct investment exerts a detrimental influence on the environment by contributing to the increase in CO₂ emissions, especially in developing countries and beneficiaries of remittances, who are in fierce competition to attract investors [72]. The contribution of foreign direct investment to environmental degradation is explained by the fact that many countries have flexible environmental regulations, the influx of investments leads to an increase in the consumption of non-renewable energy sources based mainly on fossil fuels, and investments challenge the development of infrastructure, such as roads, railways, ports, etc., which require intentional consumption of energy and resources, generating higher emissions [72]. If the environmental regulations of the host country are lax, industrialization and intensified economic activity negatively impact the environment, contributing to high greenhouse gas emissions and increased consumption of resources [74]. This type of behavior is associated with the “pollution haven hypothesis”, according to which countries with strict environmental standards can relocate their polluting activities to countries with less rigorous regulations, and thus, foreign direct investment can introduce methodologies and production methods that are unfriendly to the environment or that are energy inefficient, worsening green growth [74]. The sectoral composition of foreign direct investment is also decisive in the effect it induces on the environment. When foreign direct investment is targeted at industries that require energy and natural resources, water, deforestation, destruction of natural habitats, etc., the risk of environmental degradation increases [75].

The influences of human development on green growth are not yet fully understood and have led to complex challenges associated with the pursuit of sustainability [76]. The literature on the relationship between human development and green growth is still scarce, especially at the empirical level [77,78]. Human development could foster an environment conductive to innovation and entrepreneurship, laying the foundations for sustainability. Human capital is essential for green economic growth due to its role in providing an enabling environment, aiming to improve sustainable development goals and social well-being [78]. The positive association between human development and sustainability can be attributed to the high levels of education and health services that help shape human capital and play a vital role in driving green economic growth [79,80]. Although human development positively contributes to the promotion of green growth, the data suggest that this link is neither consistently steady nor entirely beneficial. Human development may correlate with increasing per capita carbon emissions due to the alterations in energy consumption and industrial dynamics [79]. Human development exerts significant pressure on the environment and ecology, mainly through mechanisms related to energy consumption and economic activity. Some states achieve high human development at the cost of high carbon emissions, in the so-called negative expansionary decoupling state, in which carbon emissions per capita grow much faster than human development [81].

Numerous studies demonstrate the direct correlation between urbanization and green growth, either positively or negatively [82,83,84,85]. Countries should concentrate on enacting structural reforms to technologically upgrade infrastructure and industrial complexes to address the shortcomings caused by urbanization [84]. Urbanization can significantly negatively influence inclusive green growth, indicating environmental degradation due to unplanned urban growth [85]. Environmental impact damage is associated with urbanization, if rapid urbanization processes are linked to significant increases in energy consumption (particularly from fossil fuels), unsustainable industry, ungreen transportation, or carbon-intensive residential buildings [82,86]. Even while living standards are rising, and urbanization and industrialization are progressing rapidly, there are significant sustainability problems associated with these trends. Urbanization promotes development based on the high consumption of resources, environmental pollution, and ecological deterioration [87].

The global imperative for sustainable development in the context of accelerating environmental degradation, resource depletion, and climate change underscores the crucial role of political and governance factors [43,44]. Among these, political stability—and by extension, the absence of violence and terrorism—emerges as a foundational prerequisite for fostering environmentally responsible economic development. Political stability reduces policy uncertainty—widely recognized as detrimental to green growth [88]. Political stability is considered a key dimension of good governance, characterized by the effectiveness, transparency, accountability, and inclusiveness of decision-making processes within both public and private institutions [43,53]. It reflects the ability of societies to manage resources, respond to societal needs, and implement policies effectively [43]. The World Bank, for instance, defines political stability as the perceived likelihood of political instability, including politically motivated violence [89]. A strong political environment—commonly assessed through indicators such as the absence of violence/terrorism, government effectiveness, regulatory quality, and corruption control—is theorized to enhance investor confidence, enable long-term planning, and establish favorable conditions for policy implementation [43,53,89]. Conversely, unstable governments often fail to enact policies that promote entrepreneurship and safeguard financial markets [90]. Effective governance is indispensable: political stability and institutional quality mitigate the resource curse and foster sustainability [53]. We consider governance to be a cornerstone, since transparency and accountability create trust, attract investment, and enable effective resource allocation.

3. Data and Methodology

3.1. Data and Sample Considerations

The theoretical context based on the analysis of the literature review illustrated that governmental policies and public governance are essential in facilitating the development or constraining of green growth, influencing it via multiple channels, which can be studied from an economic, social, and institutional perspective. The effect of governance in a framework marked by the implementation of artificial intelligence on green growth is explored with the data collected from the websites of the Global Green Growth Institute, Oxford Insights, the World Bank, and the United Nations Development Programme.

Table 1. Description of variables.

Code	Variable	Source
GG	Green Growth Index	[31]
AIGOV	Government Artificial Intelligence Readiness Index	[30]
HDI	Human Development Index	[91]
GDP	Gross Domestic Product (constant 2015 USD)	[92]
GOVEXP	General government final consumption expenditure (% of GDP)	[92]
FDI	Foreign direct investment net inflows (% of GDP)	[92]
URBAN	Urban population (% of total population)	[92]
POLSTAB	Political Stability and Absence of Violence/Terrorism (index)	[92]

describes the variables and data sources.

As the influence of government artificial intelligence on green growth is a relatively underexplored topic in the literature, the present study addresses a significant gap in the existing studies. The dependent variable refers to the Green Growth Index (GG) [31], and the core explanatory variable is the Government Artificial Intelligence Readiness Index (AIGOV) [30]. The sample is represented by the countries of the European Union, from 2019 to 2023.

To control for macroeconomic factors, this study uses the gross domestic product (GDP), general government final consumption expenditure (GOVEXP), and foreign direct investment net inflows (FDI). The introduction of the GDP as an explanatory element is based on the fact that expanding economies possess more public and private resources, which can be directed both towards the advancement of AI in government and towards green technologies, infrastructure, and innovation. Government expenditure is included as a proxy for the link between AI in governance and green growth, reflecting the essential role of public spending in financing projects related to AI implementation in public services, as well as in supporting green infrastructure, investments in education and human capital, and research and development oriented towards sustainable technologies. Foreign direct investment is also incorporated to capture the role of multinational companies in either promoting or hindering green growth. Its inclusion is motivated by the observation that foreign direct investment inflows often bring business based on intensive technologies in which AI and sustainability overlap. Economies that attract foreign direct investment are more likely both to integrate AI into governance and to create environments conductive to the operations of multinational enterprises.

To reflect the social dimension, this study uses the Human Development Index (HDI) and the degree of urbanization (URBAN). For the institutional dimension, the Political Stability and Absence of Violence/Terrorism indicator (POLSTAB) is employed. The introduction of the Human Development Index is justified by the idea that an educated healthy population with adequate income is a fundamental condition for the effective and ethical implementation of AI in public services, and because higher levels of human development foster technological absorption capacity, stimulate social demand for environmental protection, and reduce inequalities that may otherwise hinder the transition towards sustainable growth. Moreover, advanced human development is associated with stronger civic engagement and more capable institutions, both of which reinforce the capacity of AI-driven governance to contribute to green growth and sustainable development. Urbanization is considered another relevant factor, since a growing urban population is associated with rising demands for efficient public services, including those based on AI. Moreover, the goals to develop smart and sustainable cities inherently involve advanced digital tools, while also requiring policies to mitigate the risks of environmental degradation. Finally, political stability is included as it constitutes a precondition for consistent, long-term public policies—both those leveraging AI in governance and those towards sustainability. The prospects for green growth are strengthened in a politically stable country because it guarantees policy consistency, credibility, and effective service delivery to its population.

The Human Development Index is available from the United Nations Development Programme [91], while the rest of the control variables are selected from the World Bank [92]. The choice of these categories of control variables is based on the recent literature explaining the main determinants of green growth, which was referenced in the previous section of this study. We opted for the composite nature of these control variables by selecting both macroeconomic indicators, as well as social and state governance proxies, to capture their dominant interactions with green growth as comprehensively as possible.

The Conceptual Research Framework (Figure 1) synthesizes insights from institutional economics, technological adoption theory, and environmental governance literature. The model suggests several interdependent channels through which governmental AI influences green growth dynamics. We propose a multidimensional framework of economic, governance, and social factors, which can moderate and modulate the channels through which governments’ AI readiness influences green growth.

In contrast with the business sector, where the implementation of artificial intelligence has been more extensively studied, the challenges associated with deploying artificial intelligence in public administration remain insufficiently understood. The existing body of knowledge on artificial intelligence integration in public sector contexts is relatively limited compared to the substantial research conducted on its application in business environments [93,94,95,96,97,98]. Fortunately, in parallel with the exponential growth of artificial intelligence adoption and deployment in the public sector in recent years, the academic literature has begun to expand, addressing a wide array of issues related to governmental artificial intelligence implementation. The evolving body of research can be broadly categorized into several key thematic areas: the assessment of the current state of artificial intelligence deployment in public administration, artificial intelligence-driven policymaking, associated challenges and opportunities, determinants, technological dimensions, ethical considerations, issues of transparency and accountability, among others [94,99,100].

Despite the growing relevance of artificial intelligence in public sector governance, there remains a significant gap in the literature concerning its impact on green growth. While both artificial intelligence and green growth have emerged as critical drivers of sustainable development and economic transformation, their intersection remains underexplored. This study seeks to address this gap by focusing on the nexus between governmental artificial intelligence and green growth, recognizing their pivotal roles in environmental sustainability, resource efficiency, and socio-economic resilience. To address this gap, we estimate the effect of governments’ readiness to adopt artificial intelligence on green growth, based on the panel data from the period 2019 to 2023 for the European Union countries. The analysis of this region reveals a special group of countries that present structural heterogeneity in a common European framework and that are developed differently from an economic, social, and administrative point of view. The study of these states is based on their integration into the structure of the European community, which is deeply burdened by legal regulations that significantly influence national policy-making processes. Although these countries retain sovereignty over domestic public and governmental policies, they are obliged to harmonize these policies with the Community norms and thus acquire the prerequisites for a high degree of governmental homogeneity.

Even if there are premises for a degree of governmental homogeneity among European Union countries, the pace of implementation of artificial intelligence at the governmental level varies significantly across countries. This variation is driven by disparities in economic advancement, technological dimensions, social structure, and human capital capabilities, and diverse levels of the digitalization of government services, artificial intelligence maturity, and governance styles across countries contribute to these discrepancies [101,102,103]. Integrating AI into governance could be affected by people’s trust in the government or in AI [104,105]. Evidence of differing governmental approaches to artificial intelligence is also reflected in the highly heterogenous manner in which countries formulate and publish their national AI strategies. To date, over 60 countries, more than more than 70% of which are developed, have published their national AI policies highlighting a significant disparity and the relative lack of progress in AI policy development among low-income countries [94,106].

The limited time frame of our analysis is due to the scarcity of comprehensive and comparable data regarding the degree of artificial intelligence implementation at the government level. The relatively recent and exponential advancement of artificial intelligence technologies—particularly over the past five years—has outpaced the development of standardized monitoring and reporting mechanisms. As a result, data on the institutional readiness of states to adopt and integrate artificial intelligence into public administration remain fragmented and inconsistent. Moreover, only a limited number of international organizations and research institutions systematically collect and publish relevant indicators on artificial intelligence in governments. This data gap constrains large longitudinal analysis and requires a focus on more recent years where reliable, consistent, and harmonized data are available. Previous research highlights the limited availability of indicators that capture governments’ readiness for artificial intelligence adoption. One of the most recent composite measures in this domain is the Global Index for Responsible AI, which was first released with data for the year 2024 and covers 138 countries. The index, constrained to an integer scale from 1 to 100, assesses governments’ commitments and national capacities for the responsible development of AI, integrating technical, social, and political dimensions [107]. To enable longitudinal analysis, our study employs the Government Artificial Intelligence Readiness Index, conceptually developed in 2017. For this analysis, we use data from the 2019–2023 period, as provided by Oxford Insights [30]. The literature utilizing this index is growing, because it offers valuable insights into key governmental dimensions, including the AI-policy frameworks, regulatory structure, and strategies for AI adoption, along with technological advancements, data use, and infrastructure [108,109,110]. We have not identified other indicators that systematically capture governments’ implementation of artificial intelligence over time in a way that supports longitudinal analyses. Although there is a growing tendency to develop indices specifically targeting the governmental dimension of AI, these instruments are either recent or lack temporal depth. For example, a new index—AI Governance International Evaluation Index AGILE—has been introduced in 2024 but it covers only 14 countries [111]. Another index, the AI Preparedness Index, has been launched by the International Monetary Fund in 2023, covering 174 countries [112]. This index aggregates a broad set of macro-structural indicators across four key dimensions: digital infrastructure, human capital and labor market policies, innovation and economic integration, and regulation and ethics [112]. It provides a holistic view of a country’s readiness for AI adoption and comprises a regulation and ethics dimension, that evaluates the adaptability of existing legal and regulatory to evolving new (digital) business models and the presence of strong governance for effective enforcement [112].

Governments cannot always prioritize the allocation of public funds to major investments in technologies, equipment, and human resources supporting digitalization and the implementation of artificial intelligence, as governmental priorities may often require other types of public spending, as was the case during the COVID-19 pandemic. During this time, local governments across the European Union shifted expenditure priorities, with increased public spending on healthcare and social protection [113,114]. At the expense of investment, public expenditures—particularly those supported by EU funds—focused on sustaining healthcare systems, preserving existing businesses, and supporting individuals in the labor market [115]. Governments faced the fiscal dilemma of allocating limited resources, and consequently, public investments in equipment, technologies, and human resources essential for artificial intelligence development at the governmental level were not prioritized during this period. Investments in digital infrastructure and artificial intelligence without preparedness may hinder essential social progress and exacerbate inequalities, especially in developing economies [116]. Nevertheless, the pandemic underscores the need to integrate modern digital technologies across all sectors of economic and social life, including in the political field [117,118]. In response, the European Union launched the Digital European Programme, with a total budget exceeding EUR 8.1 billion, aimed at promoting the adoption of digital technologies by businesses, citizens, and public administrations [119].

In fact, since 2018, the European Union has developed strategic policy documents in the field of artificial intelligence, emphasizing Europe’s role as a key actor in the development of artificial intelligence-based platforms that provide services to companies and organizations, as well as for e-government systems [120]. To foster societal acceptance rather than the imposition of artificial intelligence, the governments of the EU states aimed to bring together the efforts of social partners and civil society bodies and to ensure that its benefits are widely disseminated, that all citizens are adequately prepared to make the most of this technology, and that there is a wider reflection on potential deeper societal changes [120].

A key milestone in the development of a public framework based on responsible artificial intelligence is the adoption, in 2024, of the Artificial Intelligence Act in the European Union [121]. The Artificial Intelligence Act establishes a comprehensive legal framework based on clear principles, transparency obligations for providers of artificial intelligence systems, prohibited practices, and stringent compliance requirements. The new common European framework provides a coherent and harmonized foundation for the structured, safe, ethical, and sustainable development of new artificial intelligence technologies across both public and private sectors in the European Union. In accordance with the Artificial Intelligence Act, the European Commission is mandated to develop and maintain a European Union-wide public database listing high-risk AI systems. Public authorities, agencies, or bodies intending to deploy such systems must register themselves in this database and select the system that they envisage using [121]. This unit of the EU database should be publicly accessible, free of charge, and the information should be easily navigable, understandable, and machine-readable. However, for high-risk AI systems in the area of law enforcement, migration, and asylum and border control management, the registration obligations should be fulfilled in a secure non-public section of the EU database [121]. Access to this secure non-public section should be strictly limited to the Commission as well as to market surveillance authorities regarding their national section of that database [121]. Furthermore, a public consultation on high-risk AI systems has been launched in July 2025, inviting a broad spectrum of stakeholders to contribute their views and recommendations [122]. The regulatory processes in the field of artificial intelligence are inherently complex and demanding, requiring time, meticulous attention to technical and legal details, and the sustained involvement of institutional public and civil society.

Within the wider context of the European Union’s sustainability objectives, the challenges of artificial intelligence implementation are increasing with those of the green transition and decarbonization. The ambitious climate targets and carbon neutrality by 2050 of the Green Deal [123] require coordinated strategies for the advance of green technologies, as well as those of artificial intelligence.

3.2. Model and Econometric Specification

To examine the influence of artificial intelligence on green growth in the European Union between 2019 and 2023, the basic explanatory variable of the model refers to the degree of readiness of governments for the adoption of artificial intelligence, a composite indicator developed by Oxford Insights [30], and green growth is captured through the Green Growth Index available at the Global Green Growth Institute [31].

Given the quasi-recent nature of the concepts associated with artificial intelligence and the relatively recent concerns to quantify this phenomenon, there are limited datasets that illustrate the degree of preparedness of governments for the adoption of artificial intelligence. The limited availability of data, particularly for the 2019–2023 interval, constrained the temporal dimension of the panel and, consequently, precluded the use of certain econometric models that require long panels to produce consistent and efficient estimates.

To identify the most appropriate econometric testing techniques, the data available for the European Union country panel (2019–2023) were analyzed, and a regression diagnosis was performed to test the assumptions of classical regression methods, related to multicollinearity, stationarity, cross-sectional dependence, heteroskedasticity, serial correlation, and normality [124]. Given that the classical linear model assumes strict exogeneity of regressors and that ignoring endogeneity could yield biased results [125], a preliminary analysis of potential endogeneity was undertaken using the Limited-Information Maximum Likelihood (LIML)—an instrumental variable approach [126]. However, the identification of valid and strong instruments for each potentially endogenous explanatory variable posed significant methodological challenges. Endogeneity arises when relevant variables are omitted from the model and exert a significant influence on the dependent variable, when simultaneity or reverse causality exists between variables [127], or when dynamic endogeneity is present, due to the past value of dependent variable’s influence on its current value [128].

To estimate the influence of government artificial intelligence readiness on green growth in the European Union (2019–2023), this study performs several static (Equation (1)) and dynamic (Equation (2)) panel models, based on the following specifications:

$${\mathrm{G}\mathrm{G}}_{\mathrm{i},\mathrm{t}}={\propto }_0+{\propto }_1{{\mathrm{A}\mathrm{I}\mathrm{G}\mathrm{O}\mathrm{V}}_{\mathrm{i},\mathrm{t}}+{\propto }_2\mathrm{G}\mathrm{D}\mathrm{P}}_{\mathrm{i},\mathrm{t}}+{{\propto }_3\mathrm{G}\mathrm{O}\mathrm{V}\mathrm{E}\mathrm{X}\mathrm{P}}_{\mathrm{i},\mathrm{t}}+{{\propto }_4\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i},\mathrm{t}}+{{{\propto }_5\mathrm{H}\mathrm{D}\mathrm{I}}_{\mathrm{i},\mathrm{t}}+{{\propto }_6\mathrm{U}\mathrm{R}\mathrm{B}\mathrm{A}\mathrm{N}}_{\mathrm{i},\mathrm{t}}+{{\propto }_7\mathrm{P}\mathrm{O}\mathrm{L}\mathrm{S}\mathrm{T}\mathrm{A}\mathrm{B}}_{\mathrm{i},\mathrm{t}}+\mathrm{u}}_{\mathrm{i},\mathrm{t}},$$

(1)

$${\mathrm{G}\mathrm{G}}_{\mathrm{i},\mathrm{t}}={\propto }_0+{{\propto }_1\mathrm{G}\mathrm{G}}_{\mathrm{i},\mathrm{t}-1}+{\propto }_2{{\mathrm{A}\mathrm{I}\mathrm{G}\mathrm{O}\mathrm{V}}_{\mathrm{i},\mathrm{t}}+{\propto }_3\mathrm{G}\mathrm{D}\mathrm{P}}_{\mathrm{i},\mathrm{t}}+{{\propto }_4\mathrm{G}\mathrm{O}\mathrm{V}\mathrm{E}\mathrm{X}\mathrm{P}}_{\mathrm{i},\mathrm{t}}+{{\propto }_5\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i},\mathrm{t}}+{{{\propto }_6\mathrm{H}\mathrm{D}\mathrm{I}}_{\mathrm{i},\mathrm{t}}+{{\propto }_7\mathrm{U}\mathrm{R}\mathrm{B}\mathrm{A}\mathrm{N}}_{\mathrm{i},\mathrm{t}}+{{\propto }_8\mathrm{P}\mathrm{O}\mathrm{L}\mathrm{S}\mathrm{T}\mathrm{A}\mathrm{B}}_{\mathrm{i},\mathrm{t}}+\mathrm{u}}_{\mathrm{i},\mathrm{t}}$$

(2)

where i represents the country, t is the period (years), GG_i,t−1 represents 1-year lag of GG (Green Growth Index), and the rest of the notations represent the variables described in Table 1. α₀ is a constant (intercept), α_{1,2,3,4,5,6,7,8} are the coefficients of the estimated parameters, and u_i,t is the error term.

Based on the results of the preliminary diagnostic tests, we begin with a benchmark specification using Ordinary Least Squares (OLS) regression, controlling for heteroscedasticity. To assess the robustness and sensitivity of the results, we further estimate two additional static panel models: the Panel Corrected Standard Errors (PCSE) which allows for groupwise heteroskedasticity across cross-sectional units and the Feasible Generalized Least Squares (FGLS) method, assuming the disturbances to be panel-level heteroskedastic. Subsequently, to test endogeneity concerns, for each explanatory variable, we estimate the model using the Limited-Information Maximum Likelihood (LIML) technique in a static framework [126]. This method is suitable for smaller samples and is robust to weaker instruments, making it a less biased alternative to 2SLS. Finally, we employ a dynamic panel estimation using the System Generalized Method of Moments (System GMM) estimator, which is particularly suited to effectively addresses issues of dynamic endogeneity, the omission of the variables, and simultaneous causality [129,130,131,132]. Econometric data were processed using STATA (https://www.stata.com/).

4. Empirical Results

Summary statistics of the variables are presented in

Table 2. Descriptive statistics.

Variable	Level	Mean	Std. Dev.	Min.	Max.
GG	overall	66.545	5.242	53.46	75.01
	between		5.305	53.882	74.686
	within		0.425	64.795	67.489
AIGOV	overall	66.204	8.657	47.93	88.103
	between		8.257	50.313	80.229
	within		2.966	58.482	75.861
GDP	overall	5.55 × 1011	8.56 × 1011	1.40 × 1010	3.70 × 1012
	between		8.69 × 1011	1.60 × 1010	3.66 × 1012
	within		3.46 × 1010	3.75 × 1010	6.75 × 1011
GOVEXP	overall	20.245	3.021	11.060	26.531
	between		2.953	11.867	25.898
	within		0.818	18.233	22.717
FDI	overall	10.816	78.745	−440.131	433.861
	between		42.637	−47.283	212.620
	within		66.612	−382.030	403.975
HDI	overall	0.907	0.035	0.817	0.962
	between		0.035	0.831	0.957
	within		0.005	0.893	0.921
URBAN	overall	73.976	12.931	53.729	98.189
	between		13.124	53.849	98.115
	within		0.343	72.898	75.040
POLSTAB	overall	0.673	0.248	0.101	1.333
	between		0.240	0.145	1.171
	within		0.076	0.484	0.838

Source: Authors’ processing. Note: GDP is used in its logarithmic form.

. From 2019 to 2023, the Green Growth Index in the European Union countries has been relatively stable and has changed slowly. The variability of this indicator is due to differences between countries, as the between-country variation (5.31) exceeds the within-country difference (0.43). This index had an overall mean of 66.55, with a relatively low standard deviation of 5.24. The distribution of the values between the minimum of 53.46 and the maximum of 75.01 denotes a slightly narrow spread. The degree of implementation of artificial intelligence by the governments analyzed varies substantially between countries (a standard deviation of 8.66), presenting an average of 66.20 and values in the range of 47.93 and 88.10, which reflects moderate distribution. The most important variation of this indicator is cross-sectional between countries (between-country variation is 8.26) and denotes different degrees of the readiness of governments to implement artificial intelligence. Regarding the macroeconomic situation, the analyzed indicators show substantial heterogeneity across countries and minor changes over the analyzed period, both in terms of the Gross Domestic Product and the general government final consumption expenditure. Notable differences between states can be found in terms of the Gross Domestic Product (between-country variation of 8.69 × 10¹¹), but over time, it has registered smaller fluctuations within a given country (within-country variation of 3.46 × 10¹⁰). The percentage of government spending as a percentage of GDP ranges from 11.06% to 26.53%, with a standard deviation of 3.02%. The differences between states in terms of this indicator are significant (between-country variation of 2.95%), while during the studied time interval, the values of the indicator were relatively stable (within-country variation of 0.81%). On the contrary, for foreign direct investment, the situation is different in terms of variations between years and countries, reflecting a highly volatile indicator, with a mean of 10.82 but with a large standard deviation of 78.75. This indicator registers values between a minimum of −440.13 and a maximum of 433.86, suggesting the presence of extreme outlier values. A considerable fluctuation over time within countries is observed, based on a high value of 66.61 for within-country variation. The Human Development Index values display a very stable indicator across countries and time, with a mean of 0.91, a very low standard deviation of 0.035, and a range between 0.817 and 0.962. The degree of urbanization in the analyzed states falls within the range of 53.73 and 98.19, with an average of 73.98 and a standard deviation of 12.93, which denotes different degrees of urbanization. Urbanization did not show important fluctuations in the period analyzed (within-country variation of 0.34) but showed an average level of fluctuation between states (between-country variation is 13.12). A moderate heterogeneity level of institutional country factors is obtained, based on the values of Political Stability and Absence of Violence/Terrorism indicator, which fall within the range 0.101 to 1.333. It has an average of 0.673 and a standard deviation of 0.249, with slightly greater between-country variation (0.240) than within-country variation (0.076).

Based on the descriptive statistics above, several proxies of green growth exhibit substantial heterogeneity across countries. The most notable differences between states refer to the basic explanatory variable represented by the degree of implementation of artificial intelligence by governments, which presents differences in the infrastructure, equipment, and technologies of artificial intelligence used by governments, as well as dissimilarities between specific regulations and the institutional governance of countries. There are different speeds of adaptation of the governments studied to the new challenges specific to artificial intelligence, against the background of different degrees of economic development, mentality, culture, levels of government expenditures, or the erosion of the quality of governance, by weakening the levers of the rule of law and sustaining climates characterized by political instability or / and corruption. Foreign direct investment also registers substantially different values between states, characterized by extreme values at both poles and notable differences between countries. The heterogeneity of the studied region also lies in the large ranges within which the Gross Domestic Product and governmental expenditure fluctuate, which characterizes a community area in which advanced economies display a larger Gross Domestic Product, and developing countries coexist with significantly lower absolute values of this indicator. Significant differences are also documented in the ratios of government expenditure to GDP across countries, due to a complex mix of economic, social, demographic, institutional and political factors.

Disparities between countries are lower in terms of the green growth index, whose values fall within a relatively narrow range of values, as well as in the case of the human development index, which shows a high level of homogeneity of human development between states. Also, during the analyzed period, moderate fluctuations are recorded for the urbanization and institutional governance variables of the analyzed countries.

Figure 2 illustrates the evolution of the dependent variable, green growth, together with the trajectory of the core explanatory variable capturing governmental AI across all countries in the sample.

Following the description of the variables, a set of preliminary analyses was conducted to evaluate the econometric properties of the panel data. As shown in

Table 3. Correlation matrix of the variables (multicollinearity).

Variable	GG	AIGOV	GDP	GOVEXP	FDI	HDI	URBAN	POLSTAB	VIF
GG	1.000								1.88 mean
AIGOV	0.452	1.000							3.10
GDP	0.409	0.535	1.000						1.91
GOVEXP	0.524	0.358	0.296	1.000					1.37
FDI	−0.275	−0.074	−0.207	−0.115	1.000				1.09
HDI	0.246	0.734	0.422	0.335	−0.035	1.000			2.45
URBAN	−0.062	0.506	0.200	0.432	0.099	0.524	1.000		1.67
POLSTAB	0.252	0.356	−0.182	−0.051	0.117	0.347	0.187	1.000	1.59

Source: Authors’ processing. Note: VIF is the Variance Inflation Factor.

, the pairwise correlation coefficients among the regressors are all below the commonly accepted threshold of 0.8, which suggests the absence of strong multicollinearity [133]. The highest observed correlation is between HDI and AIGOV (0.734), indicating a moderate degree of collinearity. To further assess this, the Variance Inflation Factor (VIF) was computed for each explanatory variable. The average VIF across the panel is 1.88, with all individual VIF values below 3.1. These values fall well within the acceptable limits, supporting the conclusion that multicollinearity is not a substantive concern in the dataset [134].

The stationarity of the panel variables was assessed using the Phillips–Perron (PP) unit root test [135], with the results presented in

Table 4. Results of stationarity and cross-sectional dependence.

Variable	Stationarity	CSD
	Phillips–Perron Test	Pesaran CD Test
GG	860.263 ***	11.886 ***
AIGOV	542.054 ***	0.447
GDP	403.999 ***	18.066 ***
GOVEXP	186.953 ***	14.017 ***
FDI	109.550 ***	13.309 ***
HDI	176.805 ***	21.995 ***
URBAN	372.403 ***	32.421 ***
POLSTAB	151.957 ***	6.232 ***

Source: Authors’ processing. Notes: *** denotes significance at the 1 percent level. To control for deterministic linear trend, stationarity for GDP, HDI, and POLSTAB is tested with demean and trend options.

. The null hypothesis of unit roots is rejected for most variables, indicating that they are stationary. For GDP, HDI, and POLSTAB, stationarity is achieved only around a common deterministic trend across countries, with demean and trend options in the PP test. While these options typically lead to a reduction in the test’s power, their application is methodologically justified in the presence of unobserved common factors or global shocks (accounted for by demean) and for variables exhibiting deterministic trends over time (captured by trend). The analysis period encompasses the years impacted by the COVID-19 pandemic, during which global economic shocks occurred. However, the shocks’ effects on the economy appear to have been transitory, as key macroeconomic and governance indicators followed a deterministic trajectory, sustained by coherent economic policies underpinned by structural factors [136].

Cross-sectional dependence was evaluated using Pesaran’s CD test [137,138]. As reported in Table 4, the test rejects the null hypothesis of weak cross-section dependence, for all variables except AIGOV, suggesting the presence of strong cross-sectional dependence in the majority of the series. This finding indicates that most variables are influenced by common global or regional factors, consistent with macroeconomic interdependence across countries. In contrast, the absence of cross-sectional dependence in AIGOV suggests that it is primarily driven by country-specific domestic processes, typically conducted internally, with minimal influence from common spillover effects.

To examine heteroskedasticity, both the White and the Cameron and Trivedi tests were employed [139]. The tests statistics, presented in

Table 5. Results of heteroskedasticity.

Heteroskedasticity	Stat.	p-Value
White test	57.86	0.008
Cameron and Trivedi test	67.85	0.009

Source: Authors’ processing.

, reveal significant heteroscedasticity in the residuals.

Serial correlation was tested using the Wooldridge test for autocorrelation in panel data [140,141]. The results, summarized in

Table 6. Results of serial correlation test.

Serial Correlation	Stat.	p-Value
Wooldridge test	24.902	0.000

Source: Authors’ processing.

, lead to rejection of the null hypothesis of no-first order autocorrelation, confirming the presence of serial correlation in the residuals.

Finally, the normality of the variables was evaluated using skewness and kurtosis statistics, as summarized in

Table 7. Results of normality test.

Variable	Normality
	Skewness	Kurtosis	p-Value
GG	0.060	0.918	0.164
AIGOV	0.360	0.803	0.634
GDP	0.528	0.001	0.009
GOVEXP	0.138	0.132	0.102
FDI	0.000	0.000	0.000
HDI	0.053	0.276	0.084
URBAN	0.443	0.000	0.000
POLSTAB	0.635	0.223	0.419

Source: Authors’ processing.

. Several variables (e.g., GG, AIGOV, GOVEXP, HDI, and POLSTAB) approximate normal distributions. However, others exhibit significant deviations from normality, suggesting the presence of non-normal distributions.

The baseline model was estimated using Ordinary Least Squares (OLS) with robust standard errors to address heteroscedasticity, as reported in

Table 8. Static and dynamic panel models—the effect of governmental artificial intelligence on green growth.

GG	Ordinary Least Squares (OLS)	Feasible Generalized Least Squares (FGLS)	Panels Corrected Standard Errors (PCSE)	System GMM
	Model 1	Model 2	Model 3	Model 4
L.GG	-	-	-	0.873 *** (0.035)
AIGOV	0.205 *** (0.048)	0.207 *** (0.031)	0.205 *** (0.047)	0.027 ** (0.012)
GDP	1.002 *** (0.201)	1.088 *** (0.148)	1.002 *** (0.193)	0.093 * (0.055)
GOVEXP	1.066 *** (0.124)	0.790 *** (0.083)	1.066 *** (0.121)	0.166 *** (0.040)
FDI	−0.008 *** (0.002)	−0.006 *** (0.002)	−0.008 *** (0.002)	−0.001 *** (0.001)
HDI	−29.633 *** (10.494)	−33.607 *** (6.565)	−29.633 *** (10.019)	−6.514 ** (2.638)
URBAN	−0.208 *** (0.020)	−0.188 *** (0.016)	−0.208 *** (0.020)	−0.029 *** (0.007)
POLSTAB	8.289 *** (1.211)	7.760 *** (0.723)	8.289 *** (1.127)	0.897 ** (0.452)
Constant	42.034 *** (8.517)	47.851 *** (5.577)	42.034 *** (8.125)	8.508 *** (2.927)
R-squared	0.706	-	0.706	-
Instruments	-	-	-	13
Groups	-	-	-	27
AR(1) test (p-value)	-	-	-	0.000
AR(2) test (p-value)	-	-	-	0.412
Sargan test (p-value)	-	-	-	0.411
Hansen test (p-value)	-	-	-	0.370

Source: Authors’ processing. Notes: Standard errors in parentheses; ***, **, and * denote significance at the 1, 5, and 10 percent level, respectively. Due to heteroskedasticity in the data, OLS is computed with heteroscedasticity-robust standard errors, PCSE is computed allowing for groupwise heteroskedasticity across cross-sectional units and in the FGLS method, and the disturbances are assumed to be panel-level heteroskedastic. System GMM is based on the xtabond2 twostep Stata command, with orthogonal (to use the forward orthogonal deviations transform instead of first differencing), collapse (to create one instrument for each variable and lag distance, rather than one for each period, variable, and lag distance), and robust (with Windmeijer’s finite-sample correction for two-step covariance matrix) options. The instruments are the independent exogenous variables. For AIGOV and FDI, both the contemporaneous and one-period lagged are used as instruments, owing to their high dynamic persistence, as these values typically exhibit reduced changes from one year to the next.

. To reinforce the robustness of the baseline estimates under the identified econometric issues, two static panel estimation techniques were applied under heteroskedasticity conditions, as shown in Table 8: Panel Corrected Standard Errors (PCSE) and Feasible Generalized Least Squares (FGLS). The results obtained from these estimators confirm the signs and statistical significance of the coefficients found in the OLS model, thereby validating the robustness of the baseline findings through static panel approaches. Finally, a dynamic panel model with the lag of the dependent variable is estimated, using the System Generalized Method of Moments (System GMM) method [129,130,131,132]. The endogeneity testing of the explanatory variables by LIML showed that, for each of the explanatory variables, the endogeneity hypothesis is rejected, and all are exogenous. However, the omission of some factors, unobserved heterogeneity, and autocorrelation indicate the use of a dynamic method of the GMM System type [129,130,131,132]. As part of the dynamic specification, we include the first lag of the dependent variable to capture the inertia in green growth dynamics. The estimation results are displayed in Table 8.

The estimation results underscore the robust and statistically significant impact of governmental artificial intelligence readiness on green growth across all model specifications. Additionally, the results confirm the dynamic persistence of green growth: the coefficient on the dependent variable lag1 is positive and statistically significant, reinforcing the appropriateness of the dynamic panel model and indicating a path-dependent process in green growth trajectories.

Post-estimation diagnostics support the validity of the dynamic model, system GMM. The Hansen test for overidentifying restrictions does not reject the null hypothesis, indicating that selected instruments are valid, and the model is statistically consistent. Given the concrete value of the Hansen test, there is no evidence of instrument proliferation, suggesting that the number of instruments is appropriate, and the test retains adequate power. The Arellano and Bond tests for autocorrelation show significant first-order serial correlation AR(1) and non-significant second-order correlation AR(2), thereby supporting the model’s specification and the validity of the moment conditions.

However, since the panel data originate from observational sources, the potential for endogeneity is non-negligible, which could result in biased and inconsistent OLS estimates. The limited number of years available for analysis constrains the application of methodologies relying on lagged variables to identify causal relationships, thereby posing a potential source of endogeneity. To address endogeneity issues, mitigate potential reverse causality, and unobserved heterogeneity, we employed the Limited-Information Maximum Likelihood (LIML), an instrumental variable approach [126]. Appropriate instruments were selected for each endogenous regressor based on theoretical foundations and empirical precedents from the previous literature. The results of the LIML estimation, described in

Table 9. Instrumental variable LIML—the effect of governmental artificial intelligence on green growth.

GG	Limited-Information Maximum Likelihood—Instrumental Variables (LIML-IV)
	Testing Endogeneity for:
	AIGOV	GDP	GOVEXP	FDI	HDI	URBAN	POLSTAB
	Model 5.1	Model 5.2	Model 5.3	Model 5.4	Model 5.5	Model 5.6	Model 5.7
AIGOV	0.399 *** (0.133)	0.275 *** (0.083)	0.194 *** (0.051)	0.249 *** (0.067)	0.202 *** (0.056)	0.243 *** (0.057)	0.155 ** (0.062)
GDP	0.613 * (0.359)	0.899 * (0.490)	0.974 *** (0.237)	0.938 *** (0.271)	0.998 *** (0.236)	0.922 *** (0.251)	1.287 *** (0.317)
GOVEXP	1.043 *** (0.101)	1.123 *** (0.105)	1.278 *** (0.194)	1.100 *** (0.107)	1.065 *** (0.094)	1.169 *** (0.120)	1.106 *** (0.101)
FDI	−0.007 ** (0.003)	−0.010 *** (0.003)	−0.007 *** (0.003)	−0.019 ** (0.008)	−0.008 ** (0.003)	−0.006 * (0.003)	−0.008 ** (0.003)
HDI	−48.692 *** (16.355)	−37.023 *** (12.605)	−30.949 *** (11.175)	−37.428 *** (12.770)	−28.417 * (17.167)	−22.269 * (12.992)	−35.584 *** (12.076)
URBAN	−0.227 *** (0.028)	−0.221 *** (0.028)	−0.225 *** (0.028)	−0.210 *** (0.029)	−0.208 *** (0.025)	−0.284 *** (0.063)	−0.204 *** (0.025)
POLSTAB	6.326 *** (1.748)	7.411 *** (1.906)	8.730 *** (1.311)	8.202 *** (1.541)	8.260 *** (1.278)	7.929 *** (1.296)	11.590 *** (2.748)
Constant	59.886 *** 14.552	47.292 *** (11.351)	41.372 *** (8.700)	47.532 *** (9.973)	41.274 *** (11.883)	38.738 *** (9.177)	40.026 *** (8.872)
Instruments	IIU, IIB, ISN	GEE, GNIPC, ISN	GEE, GDPPPG, UNEM, INEQINC	RQE, GDPPPG, GNIPC, UNEM	GDPPPG, CCE, UNEM, DEBT	INEQINC, ISN	INEQINC, UNEM
-Under identification test Anderson canon. corr. LM -p-value	23.139 0.000	30.703 0.000	34.728 0.000	22.854 0.000	57.567 0.000	21.629 0.000	29.045 0.000
-Weak identification test Cragg–Donald -Critical Stock–Yogo	8.627 6.46	13.012 6.46	10.737 5.44	6.509 5.44	23.047 5.44	12.030 8.68	17.270 8.68
-Overidentification test Sargan -p-value	2.808 0.245	0.721 0.697	2.139 0.544	2.266 0.519	4.870 0.181	0.165 0.685	0.130 0.718
-Over identification test Anderson–Rubin -p-value	2.838 0.241	0.723 0.696	2.156 0.540	2.290 0.514	4.960 0.174	0.165 0.684	0.130 0.718
-Endogeneity test -p-value	2.331 0.126	0.052 0.819	1.535 0.215	1.173 0.278	0.008 0.930	1.856 0.173	1.924 0.165

Source: Authors’ processing. Notes: Standard errors in parentheses; ***, **, and * denote significance at the 1, 5, and 10 percent level, respectively. Exogenous proxies used as instruments for tested exogenous variables (World Bank, 2025; United Nations Development Programme, 2025): IIU—individuals—internet use: all individuals; IIB—internet use: internet banking; ISN—internet use: participating in social networks; GEE—government effectiveness; GNIPC—gross national income per capita (2021 PPP USD), first difference; GDPPPG—gross domestic product per capita growth (annual %); UNEM—unemployment, total (% of total labor force) (modeled ILO estimate); INEQINC—inequality in income; RQE—regulatory quality; CCE—control of corruption; DEBT—claims on central government as the % GDP.

, confirm the robustness of the relationship between governmental artificial intelligence readiness and green growth, even after controlling for endogeneity. For each endogenous regressor, we associated a corresponding set of exogenous instruments (acronyms provided in Table 9). These instruments are theoretically unrelated to the dependent variable, green growth, thereby satisfying the exclusion restriction. For AIGOV, the primary explanatory variable capturing governmental artificial intelligence readiness, the instruments reflect the technological infrastructure and human behavior conductive to digital transformation: individuals—internet use: all individuals (IIU); Internet use: internet banking (IIB), and Internet use: participating in social networks (ISB). Citizens’ access to digital technologies and a higher degree of information in the social context are important prerequisites for governmental artificial intelligence implementation. For GDP, the selected instruments reveal the institutional governance of a country (captured by government effectiveness—GEE), a country’s long-term development potential (seized by gross national income per capita—GNIPC), and the diffusion of knowledge and skills across individuals and regions (internet use: participating in social networks—ISB). For the general government final consumption expenditure (GOVEXP) proxy, the configured instruments were selected to reflect the institutional governance via a country perspective and public sector dynamics: government effectiveness (GEE) captures the institutional capacity to manage public resources efficiently; GDP per capita growth (GDPPPG) affects fiscal behavior—revenue and spending decisions; unemployment (UNEM) levels can activate countercyclical public spending; and inequality in income (INEQINC) can lead to higher redistributive spending due to social protection need and pressure from vulnerable groups. For the foreign direct investment (FDI), instruments were nominated to capture the macroeconomic stability, institutional governance, and labor market fundamentals, which are determinants of a country’s attractiveness to foreign investors: the regulatory quality (RQE) indicator reflects a country’s ability to formulate and adopt adequate policies, which may indirectly affect the mechanisms through which multinational companies are attracted or not; GDPPPG provides larger domestic markets, better infrastructure, human capital, and institutions, which attract investment; gross national income per capita, in first difference (GNIPC), signals macroeconomic stability, potential larger markets, development and reliability, and predictable fiscal and monetary frameworks, which attract investors; and labor market characteristics, particularly unemployment (UNEM), affect the availability of skilled labor, critical for attracting investors. Regarding HDI, the instruments reflect macroeconomic structural development and labor market inclusion: growth in GDP per capita (GDPPPG) drives improvements in living standards over time; claims on central government as the % GDP (DEBT) influence HDI if constraining public spending on education and health; control of corruption (CCE) directly affects the quality of healthcare public outcomes, education access, and living standard through resource allocation; unemployment (UNEM) influences income and education attainment, two core components of HDI. Instruments for urbanization proxy (URBAN) were chosen based on demographic and social development determinants: inequalities in income (INEQINC) may either accelerate rural–urban migration or constrain urban consolidation; internet use: participating in social networks (ISB) could reduce or, in contrast, exacerbate information asymmetries in the urban area, influence the degree of societal cohesion, increase migration aspirations, etc. Finally, instruments for political stability, explanatory variable (POLSTAB), capture inequality and economic opportunity: inequality in income (INEQINC) can exacerbate political unrest and social instability, which may amplify or mitigate political dissatisfaction; unemployment (UNEM), as a proxy for economic disenfranchisement, is also associated with increased political instability.

5. Discussion

The empirical results demonstrate that the deployment of artificial intelligence at the governmental level has a positive and statistically significant impact on green growth across the sample countries. These findings support the acceptance of Hypothesis H₁, which posits that an increase in governmental artificial intelligence adoption is associated with enhanced green growth outcomes.

Regarding the other explanatory variables, the empirical evidence indicates that GDP growth positively contributes to green growth. These findings align with the emerging strand of literature which argues that elucidating the relationship between economic growth and environmental degradation requires datasets that incorporate both economic aspects of a country and other factors such as technology [142]. Therefore, the integration of AI into economic governance is envisioned as a catalyst for aligning economic growth with environmental sustainability by enhancing resource efficiency, stimulating eco-innovation, and reinforcing institutional capacity. While high-income economies demonstrate the possibility of absolute decoupling—where GDP expansion coincides with proportionally greater reduction in carbon intensity—this outcome is contingent upon structural transformation, technological innovation, eco-friendly approaches, and a strong policy framework [53,56,58,59]. Even though economic growth generally entails higher energy consumption, the application of AI in government can help identify early warning signals of environmental degradation, establish green priorities and regulations, and thereby contribute to the decoupling of economic expansion from environmental harm. In countries where the resources generated by rapid economic growth are allocated towards research and development, AI can further enhance green growth by supporting innovative sustainable technologies and optimizing energy systems. Moreover, in economies with high levels of income and growth, rising public concern for environmental protection exerts pressure on governments to strengthen transparency and accountability. In such contexts, AI tools integrated into the public sector can enhance the environmental performance and reinforce public trust through data-driven monitoring and reporting.

The results of this study reveal a positive relationship between government spending and green growth. The effect of government final consumption expenditure on green growth is heterogenous and strongly dependent on the composition of spending. Positive impacts materialize when resources are allocated to environmentally oriented investments, education, and innovation, while adverse effects arise when expenditure reinforces carbon-intensive structures or environmentally regressive activities [46,60,61,62,64,65]. Allocating public resources in government systems based on AI algorithms allows optimizing the use of funds for the prioritization of investments and current public expenditures towards environmentally sustainable objectives—such as conservation initiatives, renewable energy development, and the reduction in fossil fuel dependency. Sustainable public procurement and supply chains can be significantly improved through artificial intelligence algorithms that prioritize suppliers based on metrics such as carbon footprint, life-cycle greenhouse gas emissions, and compliance with environmental standards. To stimulate green growth and limit environmental degradation, government policies on the design and implementation of fiscal instruments can benefit from models simulated by governmental artificial intelligence algorithms, thus contributing to fiscal optimization processes and achieving a balance between growth and sustainability.

The findings reveal that foreign direct investment negatively influences green growth, which could be attributable to polluting technologies or industries, as well as resource extraction lacking efficient technologies, which are environmentally unfriendly to the host countries, lax settlements, the structure of foreign direct investments oriented towards industries that require resources, etc. [72,73,74,75]. Improperly configured AI government algorithms involved in approving foreign direct investment could prioritize procedural speed over environmental control. Thus, economies experience higher inflows of foreign direct investment in emission intensive domains, which negatively affects green growth. In addition, systems for monitoring the environmental impact of foreign direct investment, in which human analysis is totally excluded and which rely exclusively on government AI algorithms, are likely to report manipulated and greenwashing environmental data.

A negative relationship between the Human Development Index and green growth was shown, which may stem from the environmental pressure generated by energy consumption and activities associated with higher levels of human development. These dynamics entail the cost of elevated carbon emissions, corresponding to a state of so-called negative expansionary decoupling, in which per capita carbon emissions increase at a much faster pace than human development [79,81]. Artificial intelligence platforms, including chatbots and digital assistants, can play a critical role in enhancing environmental awareness and education among citizens [143]. These tools can disseminate information about environmental regulations, promote behavioral changes such as energy conservation and recycling, and foster a broader culture of environmental management. Further, coupling artificial intelligence systems with cutting-edge insights from the scientific literature enhances the diffusion of green innovation into public policy design and expenditure.

Evidence for the negative impact of urbanization on green growth was obtained. This could be explained by unplanned or rapid urban growth based on significant increases in energy consumption (particularly from fossil fuels), unsustainable industry, ungreen transportation, carbon-intensive residential buildings, high consumption of resources, etc. [82,85]. Artificial intelligence systems can contribute substantially to decarbonization in urban areas through applications in smart urban planning and sustainable mobility solutions. Intelligent traffic management, optimizing public transportation networks, and geospatially developing urban land-use planning are pivotal in reducing congestion, lowering emissions, and expanding green infrastructure. The predictive capacity of artificial intelligence technologies in the development of predictive risk modeling is decisive for the early management of climate-related disasters. Artificial intelligence contributes to the development of early warning systems, and the government costs associated with disaster and calamity management thus decrease considerably. Artificial intelligence technologies offer enhanced capabilities for the precise real-time monitoring and forecasting of green growth indicators. These systems reduce susceptibility to human errors, manipulation, and administrative inefficiencies. Tools such as satellite imagery and AI-driven data analytics facilitate the detection of deforestation, illegal activities, and the quantification of emissions across various sectors. Machine learning algorithms can process impressive amounts of data rapidly, enabling dynamic environmental impact assessment and evidence-based policymaking.

Further, based on the empirical results reported in Table 8 and Table 9, this study shows the positive influence of political stability and the absence of violence/terrorism on green growth. Stable countries can ensure policy continuity, which is critical for the success of green development in enhancing investor confidence, enabling long-term planning, and establishing favorable conditions for policy implementation [43,53,89]. Artificial intelligence systems improve transparency and accountability in the public field, due to specific instruments that monitor the compliance with environmental regulations. Potential cases of corruption in public procurement processes and the management of complaints about non-compliance with environmental legislation can be effectively administered through automated solutions based on artificial intelligence. The use of artificial intelligence in environmental governance also diminishes the administrative burden associated with human and technical resource deployment for environmental monitoring. Regulations targeting environmental protection are likely to be more effective and scientifically robust when they are based on substantial artificial intelligence-generated data. Furthermore, as governments consider the implementation or revision of environmentally relevant regulations, artificial intelligence systems enable comprehensive impact studies, scenario simulations, and impact assessment to guide policy formulation.

To fully leverage AI’s potential, governments must move beyond technology adoption and embrace an organization-wide readiness perspective, aligning AI initiatives with broader organizational objectives, ensuring cross-departmental collaboration, and cultivating strong managerial and political support [32]. Additionally, building a culture of innovation must be balanced with the establishment of strategic external relationships, including securing sustainable funding and designing adaptive regulatory frameworks [32]. There are opinions that identified that smaller municipalities, which often face resource constraints, can benefit from forming alliances with other entities to overcome limitations in expertise and data availability [32].

From our point of view, policymakers should prioritize investment in robust digital infrastructure, as advanced networks and computing capacity are essential enablers of AI adoption and long-term competitiveness, concurring with the existing research [144]. Moreover, we identified that national strategies must establish clear realistic objectives, complemented by incentives adapted to the diverse needs of municipalities [32]. Further, addressing the AI skills gap requires a comprehensive approach. Rather than focusing narrowly on training a small pool of experts, policies should promote nationwide upskilling and digital literacy, ensuring that citizens across society acquire at least basic AI competencies [145]. One important institutional development, regarding universities, in particular, must adapt curricula to foster AI literacy and prepare a workforce equipped with the competencies required for an AI-driven economy [35,145]. In addition, continuous professional development is essential to support reskilling throughout individuals’ careers, thereby reducing the risks of job displacement and facilitating adaptation to new opportunities [35].

6. Conclusions

This study examined the impact of governmental artificial intelligence, as measured by readiness of governments for the adoption of artificial intelligence (AI), on green growth across European Union countries from 2019 to 2023. In order to capture as accurately as possible the relationship between the variables under consideration, the study incorporates a set of explanatory factors—macroeconomic, social, and institutional governance indicators—namely economic growth, general government final consumption expenditure, foreign direct investment, human development, urbanization, and political stability. These variables are included in line with the previous literature and based on theoretical reasoning, as they represent potential moderators of the impact of governmental AI on green growth.

Our findings indicate that the explanatory variables exert distinct influences, consistent with the central research question concerning how this emerging dimension of governance, captured by governments’ readiness to adopt AI, affects green growth within the European Union. The complex and multidimensional nature of governmental AI is inherently difficult to capture. Given the pronounced dynamics of AI development and its relative recent adoption in public administration, it is understandable that only a limited number of indicators currently exist to measure AI implementation in governmental services. This study employs the Government AI Readiness Index, a composite indicator developed by Oxford Insights, as a proxy for governments’ preparedness to adopt AI. This index provides valuable insights into key dimensions of governmental capacity, including AI policy frameworks, regulatory structures, and strategies for AI adoption, alongside technological advancements, data use, and infrastructure. In this context, AI in governments reflects a combination of governance capacity (political commitment and institutional capability) and structural enablers such as technological maturity, skilled labor, data availability, and infrastructure adequacy. To capture green growth, this study utilizes the Green Growth Index developed by the Global Green Growth Institute.

The empirical results demonstrate that governments’ readiness to adopt AI exerts a consistently positive influence on green growth, thereby supporting the hypothesis that governmental AI fosters sustainable development. Furthermore, the dynamic panel model reveals evidence of dynamic persistence in green growth under the influence of governmental AI, as indicated by the positive and statistically significant coefficient of the lagged dependent variable. This confirms both the adequacy of dynamic specification and the path-dependent nature of green growth.

This research contributes to the growing body of literature emphasizing the beneficial societal effects of governmental AI, contrasting with studies that question its capacity to promote green growth—either due to concerns regarding the sustainability of AI itself or skepticism about its role in supporting sustainable development. Critics highlight the environmental implications associated with AI technologies and infrastructure, including their high energy, water, and mineral consumption, as well as electronic waste generation and the disruptive effects of automation on labor market. The main challenge for contemporary governments lies in developing and implementing responsible and ethical AI governance frameworks that serve the public interest rather than the agendas of specific groups. This study adds a crucial green dimension to the concept of governmental AI, underscoring its importance for achieving sustainable development.

The research demonstrates that the mechanisms through which the government implementation of AI contributes to green growth are multifaceted and context-dependent. The central premise of the positive impact lies in the capacity of administrative modernization, induced by AI adoption, to generate new opportunities for the advancement of the green economy. By embedding AI into economic governance, governments can create institutional and technological conditions for aligning economic expansion with sustainability objectives. This occurs through improvements in resource efficiency, the acceleration of green technology innovations and the mitigation of environmental degradation. The effect of government final consumption expenditure on green growth is heterogenous, being strongly conditioned by the structural composition of public spending. Our findings indicate that a positive impact emerges when public resources are directed toward environmentally oriented investments, education, and innovation. While government spending is biased toward carbon-intensive structures or environmentally regressive activities, the effect on green growth is negative. For the sample of countries analyzed, the results reveal that foreign direct investment exerts a negative influence on green growth. This outcome may be attributable to the prevalence of polluting technologies and extractive industries, the deployment of environmentally inefficient production processes, regulatory arbitrage due to lax environmental standards, or a foreign direct investment structure skewed toward resource-intensive sectors. The analysis further uncovers a negative relationship between the Human Development Index and green growth. This counterintuitive finding may be explained by the environmental pressure associated with higher levels of human development, such as increased energy demand, greater consumption intensity, and the expansion of activities with high ecological footprints. Such dynamics reflect the phenomenon of negative expansionary decoupling, wherein per capita carbon emissions rise at a disproportionately higher rate than improvements in human development. Additionally, the study provides evidence for the negative impact of urbanization on green growth. This is plausibly linked to unplanned or rapid urban expansion characterized by surging fossil fuel consumption, unsustainable industrialization, ungreen transportation, carbon-intensive residential buildings, heightened resource demand, etc. Conversely, the findings confirm the positive influence of political stability and the absence of violence/terrorism on green growth. Stable political environments ensure policy continuity, which is critical for the durability of green growth strategies. They enhance investor confidence, enable long-term planning, and provide favorable institutional conditions for the effective design and implementation of policies oriented to sustainability.

Policymakers should consider a multidimensional landscape of economic, institutional, ethical, and technical challenges when designing strategies to enhance AI implementation in both public service and green growth. This requires balancing innovation with accountability, efficiency with equity, and automation with human oversight. Governments must critically assess whether their administrative structures, regulatory frameworks, and human capital are adequately prepared to absorb and manage AI technologies effectively. A major priority is financing cutting-edge AI specific infrastructure for public services and training the personnel to effectively utilize these technologies. Simultaneously, aiming toward a digital culture among the broader population—one that fosters both competent use and willing acceptance of governmental AI—represents a significant challenge for current administrations. These imperatives require coordinated investment: developing technical capacity within government while building public trust and digital literacy across society. Without widespread citizen engagement and acceptance, even the most sophisticated AI systems risk underutilization or resistance, undermining their potential to drive green growth and improved public service delivery. Compliance with ethical rules and data protection represent another particularly pressing challenge, requiring robust governance mechanisms and ongoing oversight. Beyond these foundational concerns, policymakers must also address sustainability objectives, ensuring that AI deployment does not inadvertently undermine green growth through increased energy consumption or generate other environmental damages, such as electronic waste and increasing the overall carbon footprint of digital infrastructures. The ultimate goal extends beyond mere technological modernization. It is to forge a data-driven, transparent, and sustainable public administration that genuinely serves both economic prosperity, social responsibility, political stability, and green objectives.

However, this study has several limitations that may affect the robustness and reliability of its results. First, the use of the Government AI Readiness Index as a proxy for governmental AI adoption introduces a degree of subjectivity. Capturing such a complex and multidimensional phenomenon may require a comparative analysis based on another similar indicator; however, to the best of our knowledge, there is not a comparative one for the period under analysis. The scarcity of indicators that capture governmental AI is obvious. Employing alternative indices that specifically reflect the integration of AI in public services could produce different empirical outcomes. Second, the reliance on cross-sectional data from the analyzed countries, combined with their estimation using both static and dynamic panel econometric methods, may introduce certain distortion or biases in the results. The inclusion of alternative modeling techniques could enhance the reliability and internal validity of the findings. Third, the explanatory factors influencing green growth were limited to a relatively narrow set of macroeconomic, social, and institutional variables. Fourth, this study focused on European Union countries, which, while valuable, may not fully capture global variation.

Expanding the analytical framework to include additional digital determinants on green growth—such as digital infrastructure and connectivity, innovation and technology adoption, ICT investment, fintech development, socio-economic digital factors, regulatory and policy environments, etc.—could yield more comprehensive and conclusive insights. Further, future research could incorporate cultural dimensions, institutional trust and legitimacy, or/and mediating psychological mechanisms, which would act as moderators, mediators, or confounders in the empirical models. Expanding analyses to include a wider range of developing or developed countries would provide insights into how governmental AI readiness varies across different socio-economic and resource contexts.