IT University of Copenhagen, University of Stavanger, Norway & Hybrid Intelligence World

A unified regulatory taxonomy: The greatest threat to AI innovation in 2026 is a fragmented "patchwork" of laws. Success should like a consensus on core definitions what constitutes AI risk and expected impact of frontier models. If nations can align on these basics, we will avoid a "race to the bottom" where companies move to jurisdictions with the weakest safety standards. Inclusive institutional design: True success requires that the Global South is part of the lead. A win would be the establishment of a Permanent Secretariat or a global body (akin to the IPCC for climate) that provides smaller nations with the technical resources to audit AI systems without being dependent on Big Tech's self-reporting. Tangible technical benchmarks: Vague promises of Ehical AI in guidelines and Responsible AI in implmentation are difficult to enforce. A successful dialogue would produce a commitment to shared safety testing protocols. This includes: Standardized procedures and mandatory watermarking (like for synthetic media to protect democratic integrity.

• Safe, secure and trustworthy AI
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Transparency, accountability, and human oversight
• Protection and promotion of human rights

Safe, secure, and trustworthy AI This is the foundational layer. Without robust safety benchmarks, the risk of systemic bias or catastrophic failure remains high. My priority here is the transition from "black box" models to explainable AI (XAI). We need technical standards that allow users to understand why a model reached a specific conclusion, which is the only way to build long-term public trust. Social, economic, and ethical implications The "human" side of the equation is often the most volatile. Urgent action is needed to address labor displacement and the digital divide. Engagement must focus on creating "human-in-the-loop" systems that augment rather than replace human expertise, ensuring that the economic gains of AI are not concentrated within a handful of organizations or nations. Cultural and linguistic diversity AI tends to be Anglocentric because of the data it is trained on. A major priority is the development of low-resource language models such as the Projects like EuroLLM-9B and TildeOpen LLM (30B) that are optimized for European languages. For AI to be truly global, it must respect and reflect the linguistic nuances of the Global South, preventing digital colonialism where one cultural perspective dominates the information ecosystem. Interoperability and human rights As nations develop their own rules (like the EU AI Act or various Executive Orders), we risk a fractured internet. The priority here is cross-border regulatory sandboxes. We need a framework where a safe model in one country is recognized as such in another, provided it adheres to universal human rights standards, such as privacy and freedom from surveillance.

The Human-in-the-Loop mandate: We must move away from fully autonomous decision-making in critical sectors (healthcare, law, and warfare) (https://www.hybridintelligence.world). Wisdom is more than data processing; it involves empathy, context, and moral intuition-qualities that AI currently lacks. By mandating that AI serves as a Decision Support System rather than a final arbiter, we ensure that human accountability remains the ultimate authority. Value alignment: To prevent machines from outperforming us in ways that cause harm, we need intrinsic safety constraints baked into the architecture. This concerns Constitutional AI where models are trained to prioritize human rights and safety even when a harmful path might be more efficient at reaching a goal. Preserving the wisdom premium As AI masters technical tasks, our educational and economic systems must pivot to value the wisdom premium -creativity, ethics, and interpersonal leadership. That is, we must treat AI as a co-Pilot and use its speed to handle the mundane, freeing human intelligence to focus on the complex moral challenges it was meant to solve.

In the Nordic region, my work in collective and hybrid intelligence—bridging academia with my role at Hybrid Intelligence World—sits at the epicenter of a critical shift: the transition from theory to enforcement. As of May 2026, the EU AI Act has turned identified governance gaps into primary competitive benchmarks. Safety & XAI: Auditability as Currency In the Nordics, public trust is social currency. The gap in explainability is being closed by a regulatory squeeze; high-risk systems now require clear audit trails. For EduTech, Explainable AI (XAI) is no longer a "nice-to-have" but a legal mandate for market entry (Kluwer, 2026). Socio-economic: The Nordic Labor Model 2.0 Rather than fearing displacement, the region leverages labor unions to institutionalize Human-in-the-Loop systems. Governance now prioritizes competence centers that re-skill workers, ensuring AI gains are distributed rather than concentrated in global tech giants (CatalystOne, 2026). Linguistic Sovereignty: Resisting Digital Colonialism To counter Anglocentric bias, the Nordics are integrating Sovereign AI models like EuroLLM-9B. This ensures local languages and cultural nuances are preserved, preventing the erosion of Nordic identity within automated decision-making. Interoperability: The Sandbox Solution To prevent a fractured market, the Nordics lead in Regulatory Sandboxes. These frameworks allow cross-border testing, ensuring a safe model in Denmark is recognized in Finland. This interoperability is vital for scaling across the European Single Market while upholding human rights (Silent Eight, 2026). By bridging science and education with tech integration, Hybrid Intelligence World acts as a trust broker, turning these governance challenges into a blueprint for safe, secure, and trustworthy AI.

The Global Dialogue on AI Governance, established by the UN General Assembly (Resolution A/RES/79/325), should serve as the primary universal platform for harmonizing the fragmented landscape of international AI policy. Interoperability and coherence: The Dialogue should be a connective tissue between regional frameworks (like the EU AI Act), the Global South, and national strategies. By fostering regulatory interoperability, it should ensure that safety standards in one jurisdiction are recognized in others. Bridging the AI Divide: A central pillar of the Dialogue is ensuring that governance isn't dictated solely by tech superpowers. It should provide its 193 Member State, including those in the Global South, with a seat at the table. This promotes equitable access to compute and capacity-building, ensuring that global AI benefits are shared rather than concentrated. Scientific evidence-based policy: The Dialogue is supported by an independent international Scientific Panel on AI. This body should provide objective risk assessments and frontier-model evaluations, moving the conversation from political posturing to evidence-based decision-making. Human-centric norm setting: The Dialogue should institutionalize human rights, transparency, and accountability as universal benchmarks. By integrating these into a Global AI Governance Roadmap, the Dialogue ensures that automated systems align with international law and cultural diversity (linguistic sovereignty). Ultimately, the AI Dialogue transforms AI governance from a private corporate exercise into a multilateral public good, turning global frictions into a shared blueprint for safe and trustworthy innovation.

The UN High-Level Advisory Body (HLAB) on AI: Building on their recommendations for a Global AI Fund and an International Scientific Panel, the Dialogue can operationalize these concepts into treaty-level or normative agreements. OECD AI Policy Observatory: The Dialogue should leverage the OECD's repository of 650+ national AI policies to identify areas for interoperability between the EU's "Regulatory-First" model and the U.S. "Strategic-Acceleration" approach. International Network of AI Safety Institutes (AISIs): By connecting the UK, US, and EU AISIs, the Dialogue can scale technical safety benchmarks to developing nations that lack the domestic capacity to audit frontier models. Global Digital Compact (GDC): Aligning with the GDC ensures AI governance is tied to broader goals of digital public infrastructure and the Sustainable Development Goals (SDGs).

To ensure the AI Dialogue is both inclusive and actionable, it must shift from a traditional communication shop to a multi-stakeholder ecosystem. Participation should be tiered by expertise and impact, rather than just diplomatic status. Member states: Focus on harmonizing national regulations and funding Digital Public Goods. Developing nations should lead the "AI for SDGs" agenda to ensure the dialogue addresses local infrastructure gaps. Private sector: (Hyperscalers & Startups): Provide "Compute Credits" for global research and share transparency reports on model training data and safety protocols. Civil society & academia: Act as the "ethical watchdog," conducting independent audits and ensuring the protection of human rights and linguistic diversity in LLMs. Technical Standards Bodies (ISO/IEEE): Feed technical benchmarks directly into the policy stream to ensure regulations remain grounded in engineering reality.

Current AI governance is dominated by a G7+China duopoly, creating a bottleneck that excludes two critical engines of innovation: Global South nations and agile startups. The Global South: These nations are often relegated to the roles of data providers or labor pools for "ghost work" (annotation). They are often excluded from Frontier AI safety boards, leading to algorithmic colonization, where models fail to account for local legal systems, low-resource languages, or specific socio-economic contexts. Startups & Open-Source Developers: High-level discussions often risk "regulatory capture" by incumbents. Massive compliance costs designed for hyperscalers create barriers for smaller players who drive niche, "AI-for-Good" solutions (e.g. Hybrid Intelligence World), but lack the legal budgets of trillion-dollar corporations. By centering these voices, the AI Dialogue transforms from a restrictive gatekeeper into a truly global accelerator.

Participatory Regulatory Sandboxes: Instead of debating rules in a vacuum, the Dialogue should use multi-stakeholder sandboxes. Here, regulators, Global South startups, and civil society co-test emerging AI applications (e.g., decentralized data trusts) in a "safe space." This allows for experimental verification of policies before they are codified, ensuring they don't inadvertently stifle small-scale innovation. Dialogue of Dialogues & Deep Dives: The 2026 Geneva session introduced a "Dialogue of Dialogues" format, which acts as a clearinghouse for insights from other forums (G7, AU, OECD). This is complemented by Thematic Breakouts co-chaired by one Member State and one non-government stakeholder (e.g., a startup founder or academic), ensuring that "expert" silos are broken down and power is shared. Collective Intelligence Workshops: With my academic background in collective intelligence, I suggest using digital platforms for Open Consultation, the Dialogue can crowdsource "red-teaming" from global developers. This citizen science approach to policy-making surfaces lived experiences—such as how an LLM performs in an under-resourced language—that high-level diplomats often miss. The "Hub-and-Spoke" Regional Model To ensure inclusivity, the Dialogue should utilize Regional Consultative Forums (spokes) that feed into the UN Plenary (hub). By rotating these spokes through hubs like Nairobi or Bogotá, the Dialogue remains grounded in local socio-economic realities rather than just Silicon Valley interests.

Tools like Credo AI and Holistic AI automate regulatory reporting and ethical risk assessments, helping enterprises map their AI usage to specific laws like the EU AI Act. Hybrid Intelligence World works on bridging science and education with tech integration for Ethical AI and Responsible AI in organizations. We have worked with SAP and Microsoft on keynotes to teach the implementation of collective intelligence into AI model to ensure validity and saftety.