Universidade Católica Portuguesa

A successful outcome for the first Global Dialogue on AI Governance would require more than symbolic consensus; it should produce structural, normative, and operational advancements in global AI regulation. First, success would entail the recognition of AI as a challenge to the traditional architecture of international law, particularly regarding sovereignty and the distribution of power. AI systems are no longer mere instruments but increasingly autonomous and opaque actors, capable of reshaping decision-making processes and concentrating power in both States and private entities. A meaningful dialogue should therefore acknowledge algorithmic power as a new vector of geopolitical influence, requiring coordinated responses. Second, the Dialogue should lead to baseline normative convergence, even if not full harmonisation. This includes shared principles on accountability, transparency, and human oversight, particularly in high-risk domains such as military applications and automated decision-making. Without such convergence, regulatory fragmentation risks exacerbating inequalities and enabling regulatory arbitrage. Third, a successful outcome would involve institutional innovation. Existing international frameworks are ill-equipped to address the speed and complexity of AI development. The Dialogue should therefore explore mechanisms such as transnational supervisory bodies, audit frameworks, or reviewability standards to ensure meaningful oversight of algorithmic systems across jurisdictions. Fourth, the Dialogue must address technological sovereignty and asymmetries in global power. The concentration of AI capabilities in a small number of States and corporations poses risks to the autonomy of less technologically advanced countries. A successful outcome would include commitments to capacity-building, equitable access, and safeguards against digital dependency. Finally, success would mean reframing AI governance as a matter of protecting fundamental rights and preserving the rule of law in a data-driven world. This requires embedding human dignity, accountability, and legal certainty at the core of global AI governance. Anything less would be diplomatically elegant, but normatively insufficient.

• Protection and promotion of human rights
• Transparency, accountability, and human oversight
• Interoperability of governance approaches
• AI capacity-building

The selected priorities reflect a structurally grounded approach to AI governance, centred on legality, power distribution, and systemic risk. First, the protection and promotion of human rights is essential because AI systems increasingly interfere with core legal values such as dignity, autonomy, and non-discrimination. As highlighted in the underlying materials, AI is no longer merely instrumental but capable of reshaping decision-making environments, requiring a rights-based framework at the global level. Second, transparency, accountability, and human oversight are indispensable to preserve the rule of law. The opacity and complexity of algorithmic systems generate significant challenges for attribution of responsibility and legal review. Without robust accountability mechanisms, both public and private actors may exercise forms of algorithmic power that escape traditional legal constraints. Third, interoperability of governance approaches is a pragmatic necessity. Diverging regulatory models risk fragmentation, regulatory arbitrage, and geopolitical tension. Establishing minimum common standards can foster coordination while respecting different legal traditions, thereby enhancing the effectiveness of global AI governance. Finally, AI capacity-building addresses the structural inequalities embedded in the global technological landscape. The concentration of AI development capabilities in a limited number of States and corporations threatens to exacerbate dependency and undermine technological sovereignty. Supporting capacity-building is therefore crucial to ensure inclusive participation and equitable governance. Taken together, these priorities aim to move beyond declaratory principles towards a governance model that is legally grounded, globally coherent, and sensitive to power asymmetries.

Yes. While Resolution 79/325 provides a comprehensive taxonomy, it does not fully capture several cross-cutting and emerging structural issues identified in the academic literature. First, there is the issue of algorithmic power as a distinct form of geopolitical and private authority. AI is not merely a regulatory object but a vector of power concentration, particularly in the hands of a few States and large technology corporations. This raises concerns that go beyond capacity-building or human rights, touching instead on the reconfiguration of sovereignty and authority in international law. Second, an emerging issue is the crisis of legal categories and coherence. AI blurs traditional distinctions, challenging the internal logic of legal systems. This doctrinal instability risks undermining legal certainty and the intelligibility of regulation, a problem not explicitly addressed in the listed themes. Third, the political economy of data and knowledge extraction remains underdeveloped in the Resolution. Current AI systems rely on large-scale data practices that may systematically favour well-resourced actors, reinforcing asymmetries and enabling forms of exploitation that escape meaningful oversight. This goes beyond transparency concerns, requiring attention to structural imbalances in data governance. Finally, there is a need to foreground epistemic governance and the role of AI in shaping knowledge, decision-making, and legal reasoning itself. As AI increasingly participates in interpretative and predictive processes, it may subtly reshape normative outcomes, raising questions about legitimacy and democratic control. In sum, the Resolution could be strengthened by explicitly addressing AI as a transformative force on legal order, power structures, and epistemic authority, rather than solely as a set of risks to be managed.

The governance gaps identified in the selected thematic areas have concrete and asymmetrical effects on countries such as Portugal, which occupy an intermediate position, neither technological leaders nor mere passive recipients, but structurally dependent within global AI value chains. First, in terms of AI capacity-building, Portugal (and similarly positioned EU Member States) faces a structural deficit in high-performance computing infrastructure, large-scale data ecosystems, and industrial AI deployment. This limits not only innovation capacity but also regulatory autonomy, reinforcing what the literature conceptualises as technological dependency. The concentration of AI capabilities in a small number of global actors constrains the effective exercise of State sovereignty, particularly in peripheral or semi-peripheral economies. Second, regarding interoperability of governance approaches, Portugal is embedded in the EU's rights-driven regulatory model (e.g., the AI Act). While this provides normative strength, it also creates tensions with dominant market-driven or state-driven models (notably from the US and China). For countries without technological scale, interoperability gaps risk translating into regulatory dependence without technological leverage: a structurally fragile position. Third, the lack of robust transparency and accountability mechanisms at the global level disproportionately affects smaller jurisdictions. Portugal relies heavily on AI systems developed by external providers, which limits effective oversight and raises challenges for enforcement of domestic and EU legal standards, particularly in relation to opaque or proprietary systems. Finally, in the domain of human rights, governance gaps amplify vulnerabilities. Peripheral actors often become testing grounds or downstream adopters of AI systems, without meaningful participation in their design. This creates risks of imported biases, reduced democratic control, and weakened protection of fundamental rights.

The AI Dialogue can act as a platform for convergence, inclusion, and coordination in global AI governance. It enables the development of shared principles (e.g., human rights, accountability) across divergent regulatory models, reducing fragmentation. At the same time, it provides a more inclusive space, allowing countries like Portugal to participate in shaping norms otherwise dominated by major powers and tech companies. By linking political debate with scientific expertise, the Dialogue also promotes evidence-based governance. Finally, it can foster practical cooperation tools, such as common standards or oversight mechanisms, enhancing interoperability. In essence, its role is to build alignment and legitimacy in a fragmented and unequal global AI landscape.