University of Glasgow

A successful first Global Dialogue on AI Governance should deliver consensus on core principles for responsible AI, such as transparency, accountability, and fairness, while ensuring these reflect diverse, global perspectives to foster inclusivity. It must also establish actionable assurance mechanisms, like audits and certifications, to verify compliance with ethical standards in a scalable, adaptable way. Finally, the dialogue should secure commitments to include underrepresented voices, such as those from the Global South and indigenous communities, in AI governance, ensuring equitable participation and addressing global challenges collectively. Without these outcomes, AI governance risks fragmentation, lack of accountability, or failing to mitigate real-world harms, undermining trust and innovation.

• Open-source software, open data and open AI models
• AI capacity-building
• Safe, secure and trustworthy AI
• Social, economic, ethical, cultural, linguistic and technical implications of AI

From my perspective, the most urgent thematic areas for action and engagement from General Assembly Resolution 79/325 are Safe, Secure, and Trustworthy AI, AI Capacity-Building, and the Social, Economic, Ethical, Cultural, Linguistic, and Technical Implications of AI. Safe, Secure, and Trustworthy AI is a priority because the rapid deployment of AI systems demands robust safeguards to prevent misuse, bias, and harm. Without global standards for safety and accountability, AI could exacerbate inequalities, undermine trust, and pose existential risks. Active engagement here ensures that AI development aligns with human rights and societal well-being. AI Capacity-Building is equally critical, particularly for low- and middle-income countries. Bridging the AI divide requires investment in education, infrastructure, and local expertise, enabling all nations to participate in, and benefit from the AI revolution. This fosters inclusivity and prevents a concentration of power among a few tech-savvy regions. Lastly, addressing the Social, Economic, Ethical, Cultural, Linguistic, and Technical Implications of AI is essential to ensure AI systems are equitable and contextually relevant. Ignoring these dimensions could lead to cultural erasure, linguistic bias, or economic displacement. Proactive engagement in this area helps mitigate unintended consequences while maximizing AI's potential for global good. Together, these priorities create a foundation for responsible, inclusive, and sustainable AI governance, ensuring that progress benefits humanity as a whole.

Some cross-cutting and emerging issues critical to AI governance are not fully captured by the listed themes: First, AI and Climate Sustainability is a pressing concern. AI's energy-intensive models contribute to carbon emissions, while AI can also accelerate climate solutions (e.g., optimization, prediction). Governance must address AI's environmental footprint and its role in climate action, ensuring alignment with global sustainability goals. Second, AI and Democratic Integrity demands attention. AI-driven misinformation, deepfakes, and microtargeting threaten electoral processes and public trust. A governance framework should include safeguards for democratic resilience, such as transparency in political AI use and protections against manipulation. Third, AI and Labor Market Disruption requires proactive policies. Automation and AI-driven productivity shifts risk job displacement and widening inequality. Governance must anticipate these changes, supporting reskilling, social safety nets, and fair labor transitions to ensure AI benefits workers and economies equitably. These issues intersect with the existing themes but are not explicitly addressed. For example, climate sustainability ties to ethical and technical implications but needs dedicated focus. Similarly, democratic integrity and labor disruption span social, economic, and ethical dimensions but require targeted strategies. Addressing them holistically would make AI governance more future-proof, inclusive, and responsible.

In India and the UK, governance gaps in Safe, Secure, and Trustworthy AI, AI Capacity-Building, and the Social, Economic, and Ethical Implications of AI are significantly impacting the health, pharma, and digital health sectors, creating both challenges and opportunities. In India, the absence of robust AI regulations risks the deployment of biased or unvalidated health AI tools, particularly in diagnostics and drug discovery, though its vast and diverse datasets present a unique opportunity to develop inclusive AI models for global health, provided governance ensures data privacy and equitable access. Meanwhile, the UK's rapid AI adoption in the NHS, such as predictive analytics for patient care, outpaces governance, raising concerns about accountability and transparency in critical healthcare decisions; however, its leadership in pharma and drug discovery, including AI-driven advancements like protein folding, offers a major opportunity if backed by strong assurance mechanisms to maintain public trust. Cross-cutting challenges include bias in AI models, which can lead to inequitable health outcomes due to underrepresentation in training data, and data sovereignty issues, as both countries balance innovation with patient data protection under frameworks like GDPR in the UK and India's Digital Personal Data Protection Act. The greatest opportunities lie in collaborative governance, such as UK-India partnerships in digital health, to establish global standards for ethical AI, interoperability, and capacity-building, ensuring AI advances health equity rather than exacerbates disparities.

The Global Dialogue on AI Governance can serve as a catalyst for international cooperation by fostering a shared vision, trust, and actionable frameworks among nations, industries, and civil society. First, it can harmonize principles and standards, bridging gaps between existing regional regulations (e.g., the EU AI Act, US executive orders, and India's upcoming AI policy) to prevent fragmentation and create a level playing field for global AI development. By facilitating consensus on core values, such as transparency, accountability, and human rights, the Dialogue can align diverse stakeholders around a common ethical baseline, reducing the risk of conflicting or protectionist policies. Second, the Dialogue can accelerate collective problem-solving on cross-border challenges, such as AI-driven misinformation, cybersecurity threats, and bias in global datasets. For example, it could establish joint task forces or knowledge-sharing platforms to address emerging risks like deepfakes or autonomous weapons, ensuring no single country bears the burden alone. Third, it can promote inclusive participation, particularly for low- and middle-income countries, by advocating for capacity-building initiatives, technology transfers, and funding mechanisms. This would help close the AI divide and ensure that governance reflects global, not just Western, perspectives. Finally, the Dialogue can legitimize and operationalize international oversight mechanisms, such as audit protocols, certification schemes, or impact assessments, that hold both states and corporations accountable. By embedding these into trade agreements or UN resolutions, it can turn voluntary commitments into enforceable norms, making cooperation not just aspirational but practical and sustainable. In doing so, the Dialogue can transform AI governance from a zero-sum competition into a collaborative, equitable, and future-ready endeavor.

The Global Dialogue on AI Governance should build upon and connect with existing initiatives such as the OECD AI Principles, the Global Partnership on AI (GPAI), and regional frameworks like the EU AI Act and India's Digital Personal Data Protection Act, as well as industry-led efforts such as the Partnership on AI and IEEE's Ethically Aligned Design. These platforms already provide foundational principles, ethical guidelines, and technical standards, but often operate in silos or lack global consensus. The AI Dialogue could add value by bridging these fragmented efforts, creating a unified, actionable roadmap that aligns their objectives with the needs of underrepresented regions and sectors. Additionally, it could leverage partnerships like the WHO's AI for Health initiative to address sector-specific challenges, such as healthcare or climate, while ensuring inclusivity. The Dialogue's unique contribution lies in its potential to elevate political commitment, turning voluntary guidelines into binding agreements or international treaties, and establishing monitoring mechanisms to track progress. By convening a multi-stakeholder, multi-disciplinary platform, it can also amplify marginalized voices such as those from the Global South, civil society, and indigenous communities, ensuring that AI governance is not only technically robust but also socially just and globally relevant. This would transform existing initiatives from isolated efforts into a cohesive, adaptive, and equitable global framework.

• To maximize the AI Dialogue's impact, stakeholders must contribute in complementary ways: governments should align national strategies with global goals and commit to binding agreements
• industry leaders (tech, pharma, startups) can provide technical expertise, pilot ethical frameworks, and fund capacity-building in underserved regions
• civil society and academia should offer independent research, highlight risks like bias or job displacement, and advocate for marginalized groups
• intergovernmental bodies (UN, OECD, WHO) can facilitate cross-border coordination
• and youth and indigenous communities must ensure intergenerational and cultural perspectives are integrated. The Dialogue's format should be multi-layered, featuring plenary sessions for high-level priority-setting, thematic working groups for deep dives into sectors like healthcare or climate, regional hubs to tailor solutions locally, open digital forums for public input, and outcome labs to draft actionable tools like model laws or certification schemes. To ensure inclusivity and adaptability, the Dialogue should rotate hosting among regions, use hybrid (in-person/virtual) formats for accessibility, and publish transparent, time-bound action plans with clear accountability metrics, transforming it into a dynamic, globally owned process rather than a static talk shop.

Global AI governance discussions often overlook voices from the Global South, indigenous communities, rural populations, persons with disabilities, youth, and grassroots civil society, as well as smaller businesses, local researchers, and non-English-speaking experts, whose perspectives are critical to addressing AI's diverse impacts. These groups face systemic barriers, such as limited access to funding, technology, or decision-making platforms, that exclude them from shaping policies that directly affect their lives. To include them, the AI Dialogue should adopt proactive measures: First, dedicate funding and fellowships to support participation from underrepresented regions and communities, covering travel, translation, and digital access costs. Second, host regional pre-dialogues in local languages to gather input and build capacity, ensuring contributions reflect contextual needs (e.g., agricultural AI in Africa or healthcare AI in rural India). Third, mandate quotas for marginalized groups in steering committees and panels, and partner with local NGOs, universities, and community leaders to identify and amplify overlooked experts. Fourth, leverage digital tools, like multilingual platforms, asynchronous contributions, and AI-assisted translation, to lower participation barriers. Finally, prioritize intersectional perspectives by addressing overlapping issues (e.g., gender, race, disability) in AI design and deployment, ensuring solutions are inclusive by default. By structurally embedding these voices, the Dialogue can shift from a Western-dominated, tech-centric approach to a truly global, human-centered governance model.

To foster meaningful, dynamic engagement during the AI Dialogue, innovative formats should move beyond traditional panels and speeches to interactive, participatory, and immersive experiences. Gamified policy simulations could allow stakeholders to role-play as policymakers, tech developers, or affected communities, testing AI governance scenarios in real-time and revealing trade-offs. AI-powered deliberation platforms using natural language processing to synthesize and visualize diverse inputs, could enable large-scale, multilingual brainstorming, ensuring no voice is lost in translation or scale. Live red teaming exercises would invite participants to stress-test proposed AI frameworks, exposing vulnerabilities and biases in a collaborative setting. Storytelling labs could amplify underrepresented perspectives through digital narratives, art, or VR experiences, making abstract ethical dilemmas tangible. Hackathons for governance could crowdsource solutions to regulatory challenges, with winning ideas fast-tracked for piloting. Fishbowl discussions and world café sessions would break down hierarchies, allowing spontaneous, cross-disciplinary exchanges. Asynchronous "idea markets" could let participants vote on and refine proposals over time, while AI-assisted matchmaking connects stakeholders with complementary expertise or needs. Finally, hybrid listening circles blending indigenous traditions with modern facilitation could create safe spaces for marginalized voices to share unfiltered insights. By combining technology, creativity, and inclusivity, these formats would transform the Dialogue from a static event into a living laboratory for global AI governance.

Some examples of policies, practices, and platforms advancing effective AI governance could be: Policies: The EU AI Act (2024) is the first comprehensive legal framework, classifying AI systems by risk (e.g., bans on social scoring, strict rules for high-risk applications like healthcare). Singapore's AI Verify Foundation offers a testing framework for responsible AI, while Canada's Directive on Automated Decision-Making mandates transparency and human oversight in government AI use. Practices: Google's AI Principles and Microsoft's Responsible AI Standard set internal guardrails for development, including bias audits and red-teaming. The Partnership on AI (a multi-stakeholder consortium) promotes best practices like model documentation (e.g., IBM's AI FactSheets) and incident reporting (e.g., the AI Incident Database). Platforms: Hugging Face's BigScience democratizes AI development through open, collaborative model-building. AlgorithmWatch's Automating Society tracks AI's societal impact across Europe, while AI for Good Global Summit (ITU/UN) fosters cross-sector solutions for SDGs. Approaches: Participatory AI (e.g., MIT's Co-Creation Studio) involves affected communities in design, while Algorithmic Impact Assessments (e.g, the Govt of Canada AI Register) require public disclosure of AI use in government. Differential privacy (e.g., Apple's on-device processing) and federated learning enable privacy-preserving AI training