Anglia Ruskin University

1. Moving beyond the usual voices In my view, a key outcome for the first Global Dialogue on AI Governance would be its ability to move beyond the usual participants. In many of these spaces, the same actors tend to dominate discussions. A successful dialogue should make room for different perspectives, particularly from regions and groups that are often left out. It should also connect AI governance to other areas that are directly affected, such as environmental impacts, labour conditions, and access to justice; issues that are often treated separately but are closely linked in practice. 2. From principles to accountability Another important outcome would be a clearer connection between high-level principles and real accountability mechanisms. Much of the current framework relies on voluntary commitments, which has shown clear limitations. It is now evident that without enforceable standards, access to remedies remains uneven and often ineffective. A successful dialogue should engage more directly with questions of enforcement, responsibility, and the role of courts and regulatory systems. 3. Decentralising the conversation Finally, the dialogue should not remain confined to traditional global governance centres such as New York and Geneva. Expanding engagement to regional and local levels would make participation more accessible and grounded in different realities, particularly in regions like Latin America where these debates are advancing but remain underrepresented globally.

• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Protection and promotion of human rights
• Interoperability of governance approaches
• AI capacity-building

1. Protection and promotion of human rights AI systems increasingly affect a wide range of rights (not only privacy or freedom of expression). Governance frameworks need to reflect this broader impact and address risks across the full lifecycle of AI systems. 2. Transparency, accountability, and human oversight Many existing approaches rely on voluntary standards, which have shown clear limitations. There is a need for stronger mechanisms to ensure responsibility, including clearer obligations, effective oversight, and access to remedies when harm occurs. 3. Social, economic, ethical, cultural, linguistic and technical implications of AI Discussions on AI often focus on technical risks, while wider structural effects receive less attention. AI systems interact with existing inequalities, environmental pressures, and regional disparities. Addressing these dimensions is essential to avoid reinforcing or deepening those challenges. 4. AI capacity-building Capacity-building remains uneven across regions. Strengthening technical, legal, and institutional capacity is essential to enable meaningful participation in governance processes and effective implementation of standards. This is particularly relevant in contexts where regulatory frameworks are still developing.

1. Environmental impact of AI systems The environmental dimension of AI remains underdeveloped in most governance discussions. The expansion of data centres, energy consumption, water use, and extractive supply chains for hardware all raise significant environmental and human rights concerns. These impacts are often treated separately from AI governance, despite being directly linked to how systems are designed and deployed. 2. Access to justice and remedies While accountability is mentioned, there is still limited attention to how individuals and communities can actually seek remedies when harm occurs. Questions around jurisdiction, corporate structures, and evidentiary barriers make it difficult to pursue claims, particularly in transnational contexts. This is especially relevant when AI-related harms intersect with corporate activity. 3. Concentration of power and infrastructure control A small number of companies control key AI infrastructure, including compute resources, datasets, and platforms. This concentration shapes not only markets, but also governance processes and policy agendas. It raises concerns about unequal influence and limited participation in decision-making. 4. Regional inequalities and uneven participation Global discussions often overlook how uneven resources and institutional capacity affect participation. Many regions are still developing regulatory approaches but have limited opportunities to shape global standards. This risks reinforcing existing asymmetries in governance.

Governance gaps in AI are having uneven but increasingly visible effects across regions such as Latin America, as well as in sectors linked to corporate accountability and access to justice. Key challenges One of the main challenges is the gap between high-level principles and enforceable standards. While frameworks such as the UN Guiding Principles on Business and Human Rights set important expectations, implementation remains inconsistent. This limits access to remedies, particularly when harms involve cross-border corporate activity or complex supply chains. A second challenge relates to limited institutional capacity. Regulatory bodies, courts, and public institutions often lack the technical and legal resources to assess AI systems, oversee their deployment, or address disputes effectively. This creates an imbalance between those developing or deploying AI systems and those affected by them. There are also broader structural issues. AI systems can reinforce existing inequalities, particularly where data gaps, biased systems, or unequal access to infrastructure are present. At the same time, environmental pressures linked to data infrastructure (such as energy and water use) are becoming more visible but remain weakly regulated. Key opportunities At the same time, there are important opportunities. Several countries in the region are actively developing AI-related legislation and policy frameworks, which creates space to integrate human rights and accountability considerations from the outset. There is also growing engagement from civil society, academia, and public institutions, which can help shape more context-sensitive approaches.

1. Connecting fragmented initiatives AI governance is currently spread across multiple forums, with different standards and priorities. The Dialogue can help map these efforts and identify areas of overlap or inconsistency. This would make it easier to understand where coordination is needed and avoid duplication. 2. Creating shared reference points There is value in developing common expectations around key issues such as accountability, human oversight, and risk management. The Dialogue can contribute to this by clarifying how existing frameworks apply to AI systems, and where further development is needed. 3. Supporting more balanced participation International cooperation is often shaped by a limited number of countries and institutions. The Dialogue can help widen participation by creating links with regional processes and making space for perspectives that are not always present in global discussions. This is important to avoid reinforcing existing imbalances in governance. 4. Linking discussion to implementation Another role would be to connect high-level discussions with practical steps. This could include sharing experiences on regulation, oversight, and enforcement, as well as identifying common challenges faced by institutions across different contexts.

There are a number of emerging initiatives that the AI Dialogue should connect with more directly, particularly those operating outside traditional policy spaces. For example, grassroots groups and local coalitions have been documenting the environmental impact of AI infrastructure, including the expansion of data centres and their effects on water use, energy consumption, and local ecosystems. These groups often work closely with affected communities and generate evidence that rarely reaches global governance discussions. Similarly, there is a growing body of academic and practitioner-led work examining AI regulation, corporate accountability, and access to justice. Much of this work engages directly with communities affected by automated systems; whether through litigation, policy advocacy, or field-based research. However, these actors are not always included in high-level forums, where participation tends to favour a smaller group of well-established voices. This creates a gap between those shaping global narratives and those producing grounded evidence on how AI systems operate in practice. It also risks narrowing the scope of governance debates, particularly on issues such as environmental harm, labour conditions, and structural inequalities. The added value of the AI Dialogue would lie in addressing this imbalance. This means actively engaging with actors who are already working on these issues at local and regional levels, and ensuring that their knowledge informs global discussions. It also requires moving beyond a model where the same individuals and institutions are repeatedly invited to speak on behalf of broader constituencies.

Different stakeholders should contribute based on their roles and experience, but the format must allow real exchange rather than a sequence of prepared statements. Short "impulse" interventions can be useful to introduce perspectives, but they should be followed by open Q&A without restrictive pre-registration. Panels should be small (no more than 3–4 participants) and moderated to ensure interaction. Panelists should respond to each other rather than deliver prepared remarks. Without this, discussions tend to remain superficial. Greater emphasis should be placed on breakout sessions. The current balance, with a strong focus on plenaries, risks limiting meaningful participation. Increasing the number of parallel sessions would allow a broader range of stakeholders (particularly civil society and researchers) to contribute more actively. Clear outputs should also be defined in advance (e.g. identifying governance gaps or areas of convergence), so discussions move beyond general exchange.

Participation remains uneven. Governments and large companies tend to dominate visibility, while civil society organisations, grassroots groups, and smaller actors are less present. There is also limited inclusion of those working directly with affected communities, as well as researchers producing empirical and field-based evidence. These perspectives are essential but often overlooked. Inclusion should go beyond formal invitations. Speaking opportunities should be distributed more evenly, avoiding hierarchical ordering in plenaries. Mixing stakeholders across sessions (rather than grouping them by category) can improve the quality of discussion. Random or balanced allocation of speaking slots can also help reduce concentration of visibility. Practical support (financial, logistical, linguistic) is necessary to enable broader participation.

More effective engagement would come from formats that prioritise interaction and relevance. Smaller, problem-focused breakout sessions can allow deeper discussion of specific issues such as accountability gaps or environmental impacts. Country- or region-based exchanges could also be valuable, bringing together government representatives with local academics, civil society, and practitioners. This would help connect global discussions with national realities. Case-based discussions, grounded in real examples of AI deployment or harm, could further strengthen the dialogue and avoid overly abstract debates. Overall, the Dialogue should prioritise formats that enable exchange, challenge dominant narratives, and connect discussions to practical experiences.

Effective AI governance is emerging through a combination of local regulatory development and calls for stronger global coordination. At the domestic and regional level, there has been a noticeable increase in AI-related policies and legislative initiatives. While many of these remain uneven in quality, they show the importance of developing context-specific approaches that respond to local institutional capacities, social inequalities, and regulatory traditions. These processes create space to address issues that are often overlooked in global discussions, including environmental impacts, access to justice, and the role of public institutions in oversight. They also allow for experimentation and adaptation, which is essential in a rapidly evolving field. However, these developments also expose clear limitations. Fragmentation across jurisdictions, inconsistent standards, and gaps in enforcement make it difficult to ensure accountability, particularly in transnational contexts where AI systems are developed and deployed across borders. For this reason, there is increasing recognition of the need for a more coordinated global framework. Initiatives such as discussions around an international AI convention (particularly within forums like the United Nations) point to the potential for establishing shared baseline standards. A global instrument could help address cross-border challenges, clarify responsibilities, and support more consistent approaches to human rights protection and accountability.