University of Basel

A successful first Global Dialogue on AI Governance would, above all, establish trust, legitimacy and continuity in global AI discussions. Current governance initiatives are very fragmented, with different regions, actors and sectors engaging in separate forms of dialogue. It would be important to consolidate this and have a credible and inclusive foundation for international, cross-sectoral cooperation. An important outcome would be the meaningful inclusion of voices that are currently underrepresented in global AI governance debates, particularly from the Global South, smaller states, academia, civil society and technical communities. Many countries are expected to navigate the societal and security implications of AI without having equal access to compute infrastructure, datasets, expertise, or governance capacity. The Dialogue should therefore contribute to reducing these asymmetries and ensure that AI governance does not become dominated by a small number of technologically advanced actors. A second indicator of success would be the establishment of concrete mechanisms for continued exchange and coordination. This could include thematic working groups, regular expert consultations or platforms for sharing best practices on issues such as AI safety, transparency and accountability. A key aspect will be to establish a common vocabulary and shared understanding of key constituents in the debate.

• Open-source software, open data and open AI models
• Transparency, accountability, and human oversight
• Safe, secure and trustworthy AI
• AI capacity-building

The four thematic areas identified above reflect key challenges that should be prioritised in global AI governance discussions. First, safe, secure and trustworthy AI is essential to ensuring that increasingly powerful AI systems are developed and deployed responsibly. Governance efforts should address issues such as reliability, cybersecurity, misuse and societal harm in order to build public trust and reduce risks. Second, AI capacity-building is crucial to preventing growing global inequalities in access to AI technologies and governance expertise. Many countries currently lack the infrastructure, technical resources, and regulatory capacity to meaningfully participate in the AI ecosystem. International cooperation should therefore support knowledge-sharing, education, and equitable access to AI resources. Third, transparency, accountability and human oversight are fundamental safeguards. As AI systems influence more areas of public and private life, it is important to preserve meaningful human control, ensure explainability and maintain clear responsibility structures for decisions made with or through AI systems. Finally, open-source software, open data, and open AI models can help democratise access to innovation and scientific collaboration. Open approaches can support inclusivity and transparency, while also requiring careful consideration of safety and misuse risks. Together, these four themes address the interconnected goals of safety, equity, accountability and access, which are all essential for inclusive and sustainable AI governance.

One important cross-cutting issue is the growing concentration of AI infrastructure and resources. Current discussions often focus on regulating AI systems themselves, while paying less attention to the underlying distribution of compute power, data access, semiconductor supply chains, and cloud infrastructure. These structural asymmetries risk deepening existing global inequalities and limiting meaningful participation in AI development and governance. Another emerging issue is the increasing integration of AI into security, military, and information environments. AI is already shaping cyber operations, disinformation campaigns, surveillance practices, and the development of autonomous systems. These developments raise complex questions relating to international peace and security, strategic stability, and the application of international law, which deserve greater attention within global governance discussions. A further cross-cutting concern is the environmental impact of AI systems. The growing computational demands of advanced AI models have significant energy, water, and resource implications. Discussions on sustainable AI governance should therefore also consider the environmental footprint of AI infrastructure and the unequal global distribution of these costs. Finally, greater attention should be paid to the relationship between AI governance and existing international legal frameworks, including human rights law, humanitarian law, and international criminal law. AI governance should not develop in isolation, but rather in a way that complements and reinforces established international norms and principles.

In the academic and legal sector, AI is being adopted at an increasingly rapid pace, including in legal research, document analysis, predictive tools and administrative decision-making. These developments create important opportunities to improve efficiency, reduce workloads and expand access to legal information and services. AI also has significant potential to support overstretched judicial systems and facilitate more effective legal analysis. At the same time, the rapid integration of AI into legal practice and judicial environments raises profound governance challenges. Courts, prosecutors, and legal practitioners are under growing pressure to engage with AI technologies, often without sufficiently clear regulatory frameworks, technical expertise or institutional safeguards. This creates risks for fundamental procedural guarantees and human rights protections. One major concern is the lack of transparency and explainability of many AI systems. In legal contexts, decisions affecting individuals' rights and freedoms must remain understandable, contestable and subject to meaningful human review. Reliance on opaque or insufficiently validated AI tools may undermine principles such as due process, equality before the law, judicial independence and the right to a fair trial. There are also concerns relating to bias, data quality, and accountability. AI systems trained on flawed or unrepresentative data may reproduce or amplify existing societal inequalities within judicial and law enforcement processes. At present, responsibility structures for errors or harmful outcomes involving AI-assisted legal decision-making often remain unclear. These developments illustrate the urgent need for governance frameworks that ensure AI systems used in legal and judicial contexts remain human-centred, rights-compliant and subject to robust oversight. The legal sector can greatly benefit from AI, but only if technological innovation is accompanied by strong procedural safeguards and clear accountability mechanisms.

The Global AI Dialogue can play an important role as an inclusive platform for coordination, trust-building, and knowledge exchange on AI governance. Given the fragmented nature of current governance initiatives, the Dialogue could help promote greater international coherence while respecting different legal, political, and cultural approaches. A key contribution would be ensuring broader participation in global AI discussions, particularly by states with limited technical capacity, as well as academia, civil society, and technical experts. This is essential to preventing AI governance from becoming dominated by a small number of technologically advanced actors. The Dialogue could also facilitate the exchange of best practices and support cooperation on shared challenges such as AI safety, transparency, human oversight, cybersecurity, disinformation, and human rights protection. Even where binding regulation is not yet feasible, regular dialogue can help develop common norms and governance expectations. The UN is uniquely positioned to convene these discussions in a universal and multidisciplinary setting.

Different stakeholders can contribute in complementary ways. Governments can provide regulatory and policy perspectives, while international organisations can help facilitate coordination and norm development. Academic institutions and technical experts can contribute independent research and evidence-based analysis, particularly on emerging risks and technological developments. Civil society organisations play an important role in representing affected communities and safeguarding human rights considerations, while private sector actors can provide practical insights into the development and deployment of AI systems. To ensure meaningful participation, the Dialogue should be structured around thematic working groups and expert consultations focused on concrete governance challenges, such as AI safety, transparency, capacity-building, cybersecurity, and human oversight. Plenary discussions should be complemented by smaller, more interactive formats that allow for substantive exchange rather than purely formal statements. The Dialogue should also ensure balanced regional representation and provide opportunities for participation by actors from developing countries and underrepresented communities, including through hybrid participation formats and capacity-support measures.

Global discussions on AI governance continue to be dominated by a relatively small number of technologically advanced states and large private technology companies. One important gap concerns countries with limited access to AI infrastructure, compute resources, technical expertise and governance capacity. Despite being significantly affected by AI-driven transformations, many states from the Global South remain structurally disadvantaged in shaping global governance frameworks. This risks reinforcing existing digital and economic inequalities. Academic and independent research perspectives are also oftentimes overshadowed by commercial and geopolitical interests. In addition, affected communities and vulnerable groups are insufficiently represented, particularly in discussions relating to surveillance, automated decision-making, labour impacts, disinformation and the use of AI in law enforcement or conflict settings. To address these gaps, the AI Dialogue should prioritise inclusive participation mechanisms, including regional representation, hybrid participation formats, targeted capacity-building support, and dedicated opportunities for contributions from underrepresented stakeholders. Supporting equitable access to technical knowledge and policy discussions will be essential to ensuring that global AI governance develops in a genuinely inclusive manner.

One effective format could be thematic working groups or "policy labs" focused on specific issues such as AI safety, autonomous systems, disinformation, judicial AI, or digital inequality. Smaller expert-driven discussions would allow for more substantive exchange than large plenary sessions alone. Scenario-based exercises and case studies could also be particularly valuable. Discussions grounded in realistic governance challenges (for example the use of AI in elections, law enforcement, healthcare, or armed conflict) can help bridge the gap between technical, legal, and policy communities and encourage more practical dialogue. To ensure inclusiveness, hybrid participation formats should be prioritised, allowing meaningful remote participation by stakeholders who may not otherwise be able to attend. Interactive digital platforms could additionally support ongoing exchanges before and after the Dialogue itself. Another useful approach would be structured multi-stakeholder roundtables bringing together governments, academia, civil society, technical experts, and private sector actors on an equal footing. AI governance is inherently interdisciplinary, and meaningful engagement requires interaction across sectors that do not often communicate directly with one another. Finally, the Dialogue could benefit from youth engagement initiatives and opportunities for early-career researchers and practitioners to contribute perspectives on emerging technologies and long-term governance challenges. This would help ensure that discussions remain forward-looking, dynamic, and globally representative.

Regional regulatory approaches such as the European Union AI Act also provide important lessons. The Act's risk-based approach seeks to balance innovation with safeguards by imposing stricter requirements on high-risk AI applications, particularly in sensitive sectors such as law enforcement, critical infrastructure, and biometric surveillance. Open-source and collaborative initiatives also offer valuable models. Open research platforms, shared datasets, and international scientific partnerships can support transparency, reproducibility and more equitable global participation in AI development.