The special issue examines AI’s policy implications from five distinct angles. One study probes public trust and opinion within the European Union, another compares governance models in Singapore and France, a third frames global AI governance as a classic collective action problem, a fourth dissects regulatory challenges in the defense sector across the European Union and the United States, and a fifth investigates the ethical and legal implications of facial recognition technology in both democratic and autocratic states. Taken together, they map the fault lines where national sovereignty, ethical oversight, and international cooperation collide, and they suggest that no single country or institution can manage the technology alone.

The comparative analysis of national governance models reveals just how different the world’s responses have become. In a study of Singapore and France, Alibašić shows how international organizations and global governance frameworks shape national AI strategies in markedly different ways. Singapore has built an innovation-centric model, one that treats regulatory agility as a competitive advantage and leans heavily on market-driven experimentation. France, by contrast, has pursued an ethics-driven approach in which the state plays a leading role in setting boundaries and safeguarding public values. The contrast is not merely stylistic. It illustrates a fundamental tension in AI policy worldwide: whether the primary goal of regulation is to accelerate innovation or to constrain it in the name of rights, safety, and democratic accountability.

That tension becomes even sharper when the technology in question is designed for the battlefield. Parisini’s comparative study of the European Union and the United States in the defense sector underscores the fragmented nature of EU defense AI governance, which is spread across institutions and member states, and contrasts it with the far more centralized strategy of the US Department of Defense. The analysis identifies capability gaps within the EU that could be narrowed through stronger transatlantic cooperation, an argument that echoes the findings on Singapore and France about the importance of international coordination. In other words, whether the domain is civilian innovation or military capability, the same lesson emerges: jurisdictions that try to go it alone leave gaps that cooperation could close.

At the global level, the picture grows more daunting still. Olugbade situates these national differences within a broader framework, arguing that AI governance is fundamentally a collective action problem, one made worse by the non-cooperation of the three dominant powers: the United States, China, and the European Union. When the largest producers and deployers of AI cannot agree on shared rules, smaller states are left to choose among incompatible standards, and the technology flows through the cracks. Olugbade’s answer is a polycentric governance mechanism, a system in which multiple centers of authority, from national regulators to international organizations, each govern parts of the problem while reconciling national autonomy with global cooperation. Rather than waiting for a single worldwide treaty that may never arrive, polycentricity accepts overlapping jurisdictions and seeks alignment among them.

The stakes of this fragmentation are not abstract. Robles and colleagues examine regulatory frameworks for facial recognition technology across a range of political regimes and document how unevenly such measures are applied. In democratic states, facial recognition raises acute concerns about civil liberties, wrongful identification, and the chilling of lawful assembly; in autocratic states, the same technology can become an instrument of surveillance and social control. The study’s central finding, the need for harmonized international standards, echoes a sentiment running through the entire special issue: without common benchmarks, the ethical floor for AI technologies will be set by the least restrictive jurisdiction, and the consequences will spill across borders.

Public trust emerges as the connective tissue binding all of these analyses together. Jensen and Chen focus on public sentiment toward AI within the European Union and identify class-based divides in how the technology is perceived, along with skepticism rooted in deficits of governance trust. Citizens who distrust the institutions regulating AI are unlikely to accept its deployment, however technically sound the safeguards may be. The authors propose citizen engagement and transparency initiatives as mechanisms to bridge this trust gap, a micro-level prescription that complements the macro-level analyses of global governance offered elsewhere in the issue. The implication is that legitimacy must be built at both ends of the scale: in the deliberations of international bodies and in the lived experience of ordinary people confronting algorithmic decision-making.

What makes the synthesis compelling is the convergence of these five studies on a shared set of themes: governance fragmentation, public trust, ethical considerations, and international cooperation. The comparative evidence suggests that harmonizing AI governance requires multilevel collaboration, both within regions such as the European Union and across global power blocs that currently view each other with suspicion. The collective warning is stark. Without coordinated global policies, AI governance will remain uneven, and that unevenness will put democratic accountability and international security at risk simultaneously. A technology that respects no borders cannot be governed effectively by institutions that stop at them.

The policy implications are concrete. The synthesis calls for enhanced international cooperation, stronger citizen engagement to foster trust, and the establishment of ethical standards for emerging technologies such as facial recognition. It also sketches institutional mechanisms that could translate high-level policy convergence into practice without demanding complete harmonization. Regulatory sandboxes, flexible spaces where governments can trial data-sharing protocols or algorithmic audits under shared principles while maintaining oversight, offer one path for incremental alignment. International AI observatories, hosted by multilateral organizations such as the OECD, UNESCO, or the UN Tech Envoy, could coordinate monitoring, ethics benchmarking, and risk assessment across jurisdictions. Capacity-building initiatives, including South–South knowledge transfer networks and AI policy toolkits tailored to institutional contexts, would support low- and middle-income countries that currently lack the regulatory infrastructure to engage on equal terms.

The research agenda that follows is equally clear. Future work should explore mechanisms for enforcing global AI standards while respecting national sovereignty, and should conduct empirical studies on whether polycentric governance frameworks actually deliver in practice. Multi-scalar research comparing AI implementation across local, national, and regional levels would reveal where governance gaps emerge and how they might be mitigated. The special issue as a whole advocates a multi-level, polycentric approach that bridges the gap between national autonomy and global regulatory alignment, setting the stage for AI governance that is both more robust and more democratically accountable. The alternative, a world in which each bloc writes its own rules and trusts no one else’s, is not a stable equilibrium but a slow-motion race to the weakest standard. The choice, the synthesis suggests, is still open, but the window for making it deliberately is narrowing with every deployment.

Subject of Research: Comparative analysis of global AI governance frameworks and polycentric policy coordination

Article Title: From fragmentation to polycentricity: a comparative synthesis on AI governance and global policy coordination

Article References: Woods, D. (2025). From fragmentation to polycentricity: a comparative synthesis on AI governance and global policy coordination. Global Public Policy and Governance, 5(2), 87-92. https://doi.org/10.1007/s43508-025-00118-8

Image Credits: AI Generated

DOI: 10.1007/s43508-025-00118-8

Keywords: artificial intelligence, AI governance, polycentric governance, global policy coordination, public trust, facial recognition technology, EU regulation, defense sector, collective action, regulatory sandboxes, international organizations, comparative public policy

The study rests on a deceptively simple premise drawn from democratic theories of policy implementation: laws and regulations succeed more often when the administrators who carry them out understand the distinctive characteristics and demands of the governed. Applied to artificial intelligence, this means the EU’s ambitious AI agenda — including its landmark regulatory framework — depends on knowing which citizens view the technology with suspicion and why. If large segments of the public distrust AI, or distrust the institutions deploying and regulating it, compliance becomes harder, public resistance grows, and the promised benefits of first-mover adoption may never materialize.

To answer the question, the researchers turned to Eurobarometer 95.2, a survey conducted in 2021 that measured European citizens’ knowledge of and attitudes toward science and technology during the COVID-19 pandemic. Eurobarometer surveys are the European Commission’s standard instrument for gauging public opinion across member states, and the 95.2 wave included detailed questions about how respondents perceived the impacts of AI on society and the economy. The dataset and all question wording are publicly available, allowing the authors to test associations between demographic characteristics, institutional attitudes, science engagement, and AI perceptions across the Union.

The central findings are striking in their clarity. Individuals who hold positive views of EU-level governments tend to hold more favorable views of artificial intelligence. This association matters enormously in the European context, because it is precisely the EU institutions — the Commission, the Parliament, and the member-state regulators acting under EU law — that serve as the primary AI regulators. A citizen who is skeptical of Brussels is, in effect, skeptical of the referee assigned to police the technology. The study thus links a well-documented phenomenon in European politics, the persistence of Euroscepticism, directly to the emerging politics of algorithmic governance.

Equally significant is the role of social class. The researchers found that people in higher social classes tend to have more positive perceptions of AI’s impacts. This class gradient echoes a long tradition in social science research showing that policy attitudes are stratified by socioeconomic position, and it raises uncomfortable questions about who stands to benefit from the AI transition. If working-class and lower-income Europeans view artificial intelligence with greater suspicion — perhaps because they perceive greater risks to their jobs, their communities, and their autonomy — then a top-down regulatory strategy crafted largely by and for educated elites risks deepening existing divides rather than bridging them.

The third pillar of the study concerns engagement with science and technology. Respondents who attended science museums and public meetings about science were associated with more favorable views of AI, as were those who expressed a preference for involving non-experts in scientific decision-making. This latter finding is particularly intriguing: it suggests that people who believe ordinary citizens should have a voice in science policy are not reflexively anti-technology. Instead, participation itself appears to breed confidence. Exposure to scientific institutions and ideas, whether through a museum visit or a town-hall discussion, seems to demystify technologies like AI and soften public apprehension.

Taken together, these findings expose a dual trust deficit that threatens the implementation of existing EU AI policies. On one side stands a class divide, in which the populations most likely to be affected by automation and algorithmic decision-making express the least enthusiasm for the technology. On the other side stands an institutional divide, in which citizens who distrust the EU and national governments extend that distrust to the AI systems those governments are charged with overseeing. The authors argue that both deficits pose genuine problems for policy implementation, because democratic legitimacy is not merely a matter of passing well-designed legislation — it requires that the governed accept and support what is being done in their name.

The remedy the study proposes is procedural rather than technological. By transparently engaging citizens in meetings on the topic of AI, and by including non-experts in the development of AI use cases, governments may be able to improve public views of the technology. This recommendation aligns with a growing body of research on transparency and trust, which has found that explaining how automated decisions are made can increase the perceived trustworthiness of algorithmic systems, and that transparency in government generally strengthens citizens’ confidence in public institutions. It also resonates with the citizen-science movement, which has demonstrated that involving laypeople in genuine research activities builds both scientific literacy and goodwill toward scientific institutions.

The stakes of getting this right extend well beyond Europe’s borders. The EU has positioned itself as a global standard-setter in AI governance, and its regulatory approach is already influencing lawmakers in other jurisdictions. But regulatory leadership is fragile if domestic publics are not on board. Previous technology policy debacles — from troubled government website rollouts to public backlash against automated decision systems in welfare administration — illustrate how quickly implementation can falter when citizens perceive systems as opaque, imposed, or unaccountable. The study’s findings suggest that the EU’s first-mover advantage depends as much on sustained public engagement as on legislative craftsmanship.

For the scientists and policymakers watching this space, the message is clear: artificial intelligence governance is not only a technical or legal challenge but a social one. The people least trusting of AI in Europe are those with less confidence in government and lower social standing, and the most promising path to winning them over runs through transparency, participation, and genuine inclusion of non-expert voices in shaping how AI is used. If the EU can close these trust deficits, it may yet convert its regulatory ambition into democratic legitimacy — and if it cannot, the first-mover advantage it seeks may prove elusive, no matter how sophisticated its rulebook becomes.

Subject of Research: Public trust in artificial intelligence and its implications for EU AI policy implementation

Article Title: Governing AI in the EU: public perception, policy dilemmas, and the path forward

Article References: Jensen, C., & Chen, C. (2025). Governing AI in the EU: public perception, policy dilemmas, and the path forward. Global Public Policy and Governance, 5(2), 162-185. https://doi.org/10.1007/s43508-025-00114-y

Image Credits: AI Generated

DOI: 10.1007/s43508-025-00114-y

Keywords: artificial intelligence, European Union, public trust, AI regulation, Eurobarometer, social class, Euroscepticism, science engagement, policy implementation, citizen participation, algorithmic governance, public perception