Observatório Nacional de Cibersegurança e Inteligencia Artificial para a Educação (ONCIAE)

• In our view, the first Global Dialogue on AI Governance will only be truly successful if it breaks with the now too common tendency to discuss artificial intelligence at an abstract level, disconnected from the real conditions of implementation. Success will not lie in yet another declaration of principles, but in the ability to generate concrete, testable, and replicable commitments. First and foremost, we believe it is essential to recognize that AI governance cannot continue to be treated as a standalone matter. In practice, the risks do not arise from AI alone, but from its interaction with cybersecurity weaknesses, lack of oversight, deficits in digital literacy, institutional vulnerabilities, and limited operational capacity to monitor impacts. That is why, within the ONCIAE framework, we believe that real progress depends on adopting a hybrid approach that brings together AI, cybersecurity, and digital governance, particularly in sensitive sectors such as education. For this dialogue to achieve genuine historical relevance, it should deliver three very concrete outcomes: first, an international minimum framework for the responsible use of AI in educational settings, including security, data protection, human oversight, content integrity, and the protection of children and young people
• second, the launch of real pilot projects in schools, universities, and training ecosystems, where these principles can be tested through maturity indicators, auditing processes, teacher capacity-building, and incident response mechanisms
• third, the creation of a permanent cooperation network bringing together observatories, regulators, educational institutions, academia, and technology operators, capable of translating political debate into operational tools. In short, the first Global Dialogue will succeed not when it produces the most elegant text, but when it creates the conditions for territories, schools, and institutions to demonstrate, within a year, that AI can be governed with responsibility, security, and public value.

• Safe, secure and trustworthy AI
• AI capacity-building
• Transparency, accountability, and human oversight
• Protection and promotion of human rights

We selected these four priorities because they best reflect our commitment to responsible, human-centred, and operationally credible AI governance. Safe, secure and trustworthy AI is a central priority because AI systems must be reliable, resilient, and designed to minimise harm, particularly in sectors with high social impact. AI capacity-building is equally urgent, as meaningful participation in the AI ecosystem depends on strengthening institutional, technical, and human capabilities across organisations and communities. Transparency, accountability, and human oversight are essential to ensure that AI systems remain understandable, contestable, and subject to responsible governance rather than opaque or purely automated decision-making. Protection and promotion of human rights is a foundational priority, since the development and use of AI must remain aligned with human dignity, non-discrimination, inclusion, privacy, and democratic values. Taken together, these four areas combine technical robustness, institutional preparedness, ethical governance, and normative safeguards. In our view, urgent action on AI should not be limited to innovation alone; it must also ensure that deployment is safe, that institutions and people are prepared, that oversight remains meaningful, and that fundamental rights are protected. This combination of priorities therefore represents the most coherent basis for active engagement, practical cooperation, and long-term contribution to inclusive and trustworthy AI governance.

Yes. Beyond the listed themes, we believe several cross-cutting issues deserve more explicit attention. First, territorial implementation capacity should be recognised as a priority in its own right. The AI challenge is not only regulatory or ethical; it is also institutional, operational, and place-based. Responsible AI depends on the real maturity of organisations, the training of professionals, and the existence of practical mechanisms for supervision, evidence, and coordination. Second, the connection between AI, cybersecurity, and institutional resilience should be more clearly reflected. In the European context, the implementation of NIS2 creates a concrete opportunity to connect AI governance with risk management, supply-chain security, accountability, and institutional preparedness, especially in sensitive sectors such as education, public administration, and essential services. Third, the central role of the European Union should be highlighted not only as a regulator, but also as a global reference point for trustworthy, rights-based, and operational AI governance. In this context, Portugal can play a relevant bridging role between Europe, the Lusophone community, and the broader Ibero-speaking space. It is important to distinguish these two linguistic spheres: the Lusophone world and the wider Ibero-speaking space, combining Portuguese and Spanish. Through the work of the OEI, this shared linguistic and cultural space represents a strategic platform connecting education, science, culture, and international cooperation across a community of around 850 million speakers worldwide. These dimensions make the debate more practical, more inclusive, and better aligned with real implementation on the ground.

In our context, the main challenge is no longer the absence of discussion on artificial intelligence, but the disconnect between legislation and practical implementation on the ground. Regulatory frameworks and guiding principles are being developed, yet many institutions still lack the technical, organisational, and human capacity to translate them into concrete mechanisms for supervision, evidence, security, and accountability. At the same time, public and institutional debate is often captured by two equally reductive narratives. On one side, there is an overly commercial approach, focused almost exclusively on market opportunities, productivity, and monetisation. On the other, there is a defeatist and near-apocalyptic reading, presenting AI only as a threat. Neither approach is sufficient. What is needed is a balanced model of governance that is operational, territorially grounded, and institutionally credible. Another emerging challenge concerns the linguistic and cultural dimension of AI. The expansion of large language models is creating asymmetries within both the Lusophone and wider Ibero-speaking spaces. In particular, there is a growing dominance of Brazilian Portuguese over other variants of the Portuguese language, especially the one spoken in Portugal. This is not merely a linguistic issue: it affects knowledge production, education, technological mediation, cultural inclusion, and the fair representation of different language communities. At the same time, important opportunities exist. The implementation of NIS2, the regulatory centrality of the European Union, and the role of Portugal as a bridge between Europe, the Lusophone world, and the wider Ibero-American space can help foster a more integrated, secure, and culturally balanced approach. The central challenge is to turn principles into real implementation capacity, without commercial reductionism, alarmism, or linguistic erasure.

The AI Dialogue can serve as a vital multilateral instrument for fostering international cooperation on AI governance in a manner that is inclusive, balanced, and firmly grounded in the principles of the United Nations. Its foremost contribution lies in providing a trusted space where States, institutions, and relevant stakeholders may work towards **shared understanding without imposing uniformity**. In a field marked by significant asymmetries of capacity, differing legal traditions, and diverse cultural and linguistic realities, the Dialogue can help shape a common horizon of responsibility while respecting legitimate diversity in national and regional approaches. It may also play a decisive role in bridging the persistent gap between "normative frameworks and practical implementation". Many countries and sectors are not lacking in principles, but in institutional readiness, technical capability, and mechanisms for effective coordination. The Dialogue can therefore support cooperation not only at the level of ideas, but also through the exchange of practices, capacity-building, and implementation-oriented partnerships. Equally important, the AI Dialogue can help prevent the emergence of **hegemonic spaces of technological or regulatory overreach**. International cooperation should not result in new forms of exclusion, dependency, or cultural erasure. Rather, it should affirm a model of governance that protects human dignity, values pluralism, and ensures that all regions, languages, and knowledge communities can participate meaningfully in shaping the future of AI. In this spirit, the Dialogue has the potential to advance a more coherent international order for AI: one that recognises our common human foundation, while drawing legitimacy and strength from the diversity of our societies. Such an approach is essential if AI governance is to be not only effective, but also just, credible, and genuinely universal.

The AI Dialogue should build upon existing international, regional, and sectoral frameworks that have already generated valuable principles, standards, and cooperation mechanisms. In particular, it should connect with the UNESCO Recommendation on the Ethics of Artificial Intelligence, the OECD AI Principles, and, in the European context, the AI Act and related implementation mechanisms. These initiatives already provide important foundations on ethics, human rights, transparency, accountability, and trustworthy AI. It should also engage with regional and multilingual cooperation platforms that can help translate global principles into practical action. In this regard, organisations such as the OEI are especially relevant, as they connect education, science, culture, and institutional cooperation across the Ibero-American space. This is important because effective AI governance requires not only global norms, but also regional channels for implementation, capacity-building, and knowledge exchange. The added value of the AI Dialogue lies not in duplicating what already exists, but in creating a multilateral space of articulation and convergence. It can help connect normative frameworks with implementation realities, link global discussions with regional experiences, and reduce fragmentation across governance approaches. The Dialogue can also add value by giving greater visibility to the concerns of countries, institutions, and communities that are often underrepresented in global technological debates, including those facing capacity constraints or linguistic asymmetries. In this sense, it can promote a more inclusive and balanced form of cooperation, where diversity is recognised as a source of legitimacy rather than an obstacle to coordination. Its distinctive contribution, therefore, should be to act as a bridge: between principles and practice, between regions and institutions, and between our shared human responsibilities and the diversity of contexts in which AI is being developed and deployed.

In our view, the AI Dialogue should also consider the creation of a more structured diplomatic architecture for the governance of artificial intelligence and cybersecurity. Today, threats to sovereignty, institutional stability, and even human security emerge from multiple sources, many of them digital, transnational, and difficult to address through traditional diplomatic channels alone. For this reason, it may be worth exploring the establishment of an international coordination mechanism, alongside the designation by Member States of specialised diplomatic representatives or ambassadors for AI and cybersecurity. Such a model could strengthen continuity, strategic coordination, and the ability to connect foreign policy, digital governance, security, human rights, and scientific cooperation. The diplomacy required by our time must be able to speak a new language: one that understands not only borders and treaties, but also algorithms, infrastructures, cyber risks, technological dependencies, and the global asymmetries of digital power. In this sense, the AI Dialogue could help catalyse a new form of diplomacy suited to the realities of the twenty-first century.

The issue is not only that certain voices are underrepresented, but also that there are still insufficiently concrete mechanisms to include them in a stable, preventive, and decision-relevant way. If the objective is to avoid a hegemonic model of AI governance, shaped only by the largest technological or regulatory powers, then inclusion must become structural, continuously monitored, and operational. Several practical measures should therefore be considered. First, each Member State could designate an Ambassador for AI and Cybersecurity, supported by a multidisciplinary team connecting diplomacy, regulation, science, education, cybersecurity, and fundamental rights. Second, these representatives could be linked through a permanent international network, connected to the AI Dialogue, with regular meetings, early-warning exchanges, and the sharing of good practices and emerging risk assessments. Third, the Dialogue should establish a continuous monitoring mechanism, producing periodic reports on capacity asymmetries, linguistic risks, underrepresented communities, territorial impacts, and technological dependencies. Fourth, universities, observatories, governments, legislators, and territorial authorities should be integrated into permanent working structures, rather than being consulted only occasionally. Fifth, regional panels on prevention and implementation should be created to identify risks early and reduce collateral harms before they become entrenched, especially in education, public administration, essential services, and language mediation. Sixth, multilingual consultation and documentation mechanisms should be mandatory, so that global discussions are not captured by only a small number of dominant languages or language variants. Finally, the Dialogue should adopt a set of monitoring indicators measuring not only innovation, but also linguistic balance, territorial inclusion, institutional maturity, supervisory capacity, and the prevention of collateral damage. The priority is not to react only after harms occur, but to build a diplomatic, institutional, and technical infrastructure of prevention, vigilance, and continuous correction.

The most effective and innovative engagement formats should move beyond one-off meetings and create permanent structures for evidence-building, prevention, monitoring, and follow-up. In this regard, observatories can play a particularly important role. An observatory should not be seen merely as a passive space of observation, but as an institutional infrastructure for applied knowledge. It can produce evidence, identify emerging risks, monitor impacts over time, and translate technical knowledge into public policy recommendations and legislative proposals. In the field of artificial intelligence, this function is especially valuable, since governance cannot rely only on abstract principles; it requires continuous reading of reality, anticipation of risks, and adaptive institutional responses. For this reason, the AI Dialogue would benefit from formats that systematically integrate observatories, universities, governments, legislators, and regional organisations. One useful model would be the creation of permanent thematic observatory tracks, tasked with producing evidence briefs, early-warning notes, and regular implementation reports. Another relevant format would be the establishment of regional monitoring and prevention hubs, able to capture territorial specificities and connect global discussions with local and sectoral realities. Observatories can also play a decisive role in extending the Dialogue across geographies, acting as structured bridges between Europe, the Global South, the Lusophone world, and the wider Ibero-American space. In doing so, they can support a more balanced circulation of knowledge, reduce participation asymmetries, and strengthen more inclusive forms of international cooperation. The added value of such formats lies in making the AI Dialogue more continuous, more evidence-based, more preventive, and more implementation-oriented. Rather than functioning only as a forum for exchange, it can become a living mechanism for monitoring, policy learning, legislative reflection, and constructive international cooperation.

A promising good practice for AI governance is to begin with critical vertical sectors, organised in successive layers of risk, and only then consolidate broader horizontal frameworks. Rather than applying the same governance model to every sector from the outset, this approach makes it possible to prioritise those domains where the potential for systemic, social, or institutional harm is highest, such as education, health, public administration, energy, mobility, and essential services. This is consistent with the European Union's risk-based approach under the AI Act. () This method can be understood as an "onion-layer" model: first the core of highest criticality, then the outer layers of lower exposure. Its value lies in allowing regulation, supervision, and capacity-building to be calibrated according to the sensitivity of each context, thereby avoiding both fragmented resource allocation and overly abstract governance solutions. At the same time, the implementation logic of NIS2 shows that technology governance must be connected to cybersecurity, institutional resilience, and supply-chain security, which further supports beginning with the most critical sectors. () From this vertical work, it then becomes possible to derive horizontal commonalities: shared principles of accountability, human oversight, monitoring, rights protection, evidence requirements, and prevention mechanisms. In this architecture, observatories and universities can play a central role by producing applied evidence, early-warning notes, implementation reports, and legislative recommendations. At the normative level, UNESCO's Recommendation on the Ethics of Artificial Intelligence provides a human-centred framework grounded in dignity, transparency, accountability, and oversight, while organisations such as the OEI can help adapt these principles to educational, linguistic, and territorial realities across the Ibero-American space.