International Foundation for Electoral Systems (IFES)

A successful first Global Dialogue on AI Governance would be defined by tangible progress in bridging global AI governance principles with the specific realities of high-stakes domains such as elections. First, success would include the establishment of dedicated, issue-specific working structures, such as a Working Group on Elections and Election Administration. This would signal that the Dialogue is responsive to sectors where AI has distinct societal impacts and cannot be adequately addressed through purely sector-neutral approaches. Second, meaningful outcomes would involve bringing together diverse stakeholders—public authorities, election management bodies, civil society, and technology actors—into sustained and structured dialogue. The Dialogue should not only convene participants, but enable ongoing coordination and mutual understanding across institutional and technical communities. Third, success would be reflected in the translation of high-level principles (e.g., transparency, accountability, safety, and human rights) into practical guidance. This includes actionable approaches for procurement, oversight, auditability, and responsible AI deployment in real-world institutional contexts. Fourth, the Dialogue should generate greater clarity around governance gaps, including challenges related to vendor opacity, institutional dependence, and platform accountability, particularly in time-sensitive environments like elections. Finally, a successful outcome would ensure that AI governance remains grounded in societal outcomes, including strengthening public trust, safeguarding institutional neutrality, and expanding inclusive civic participation. In sum, the first Global Dialogue would be successful if it moves beyond general principles to operational, inclusive, and context-sensitive governance pathways that can be implemented and sustained across diverse global contexts.

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

IFES' selection of priority thematic areas reflects the need to ensure that AI governance frameworks are responsive to the distinct, high-stakes dynamics of elections, where technical performance, institutional integrity, and public trust are deeply interconnected. First, safe, secure and trustworthy AI is essential given the time-bound and high-visibility nature of elections. The deployment or misuse of AI systems can have immediate consequences for electoral credibility, requiring governance approaches that prioritize reliability, auditability and risk mitigation in real-world electoral environments. Second, the social, economic, ethical, cultural, linguistic and technical implications of AI are particularly pronounced in elections, where AI systems shape how citizens access information, participate in democratic processes and evaluate the legitimacy of outcomes. These impacts extend beyond technical considerations to broader societal effects, including voter behavior, information environments and institutional trust. Third, the protection and promotion of human rights is central, as elections are a core component of the right to self-determination. AI systems deployed in electoral contexts must uphold principles such as non-discrimination, inclusion and equal participation, particularly for marginalized groups and voters facing informational or accessibility barriers. Fourth, transparency, accountability and human oversight are critical given the increasing reliance of election management bodies on complex, often vendor-controlled AI systems. Governance frameworks must address risks related to opacity, limited auditability and shifting decision-making authority, ensuring that public institutions retain control and can effectively oversee AI-enabled processes. Together, these priorities reflect a focus on ensuring that AI governance is not only principled, but actionable in contexts where societal trust, institutional legitimacy and rights protection are paramount.

Yes. Beyond the listed themes, several important cross-cutting and emerging issues merit more explicit attention, particularly when considering high-stakes, institutionally sensitive domains such as elections. First, there is a clear need for context- and sector-specific governance. General thematic categories do not fully capture how AI operates within domains that have distinct legal mandates, operational timelines, and accountability structures. In elections, for example, AI systems must function within rigid timeframes and highly scrutinized processes, where even small failures can have immediate societal consequences. This highlights the importance of tailoring governance approaches to institutional contexts, rather than relying solely on sector-neutral frameworks. Second, institutional authority and control over AI systems emerges as a critical issue. Public institutions are increasingly dependent on complex, externally developed technologies, which can limit their ability to understand, audit, or meaningfully oversee these systems. Over time, this can create structural dependencies and shift decision-making influence away from accountable public bodies toward private actors or automated processes, raising governance concerns that extend beyond traditional notions of transparency or accountability. Third, there are important market-level dynamics, including opacity in AI-enabled technology markets and asymmetries between AI vendors and oversight authorities. These dynamics shape what systems are available, how they are deployed, and how much oversight is possible, yet they are not fully captured by existing thematic areas. Finally, an emerging issue is the need for structured and sustained multistakeholder coordination, particularly among public institutions, private sector actors, and civil society. In many contexts, engagement remains reactive, rather than institutionalized, and fails to engage a wide range of relevant stakeholders, limiting the effectiveness of governance efforts.

Governance gaps and ongoing advances in AI are already having significant, tangible effects on the electoral sector, creating both urgent challenges and meaningful opportunities. A central challenge stems from the rapid deployment of complex AI systems without corresponding governance frameworks tailored to electoral contexts. Election management bodies (EMBs) are increasingly required to oversee technologies they do not fully control or understand, particularly when systems are developed by external vendors. This creates risks related to limited transparency, constrained auditability, and weakened procurement oversight, all of which can undermine institutional authority and public confidence. Closely related is the challenge of growing dependency on private technology providers. Over time, this can lead to structural reliance on specific vendors or systems, potentially shifting decision-making influence away from neutral public institutions and raising concerns about long-term institutional independence and accountability. Another major gap concerns the role of AI in shaping electoral information environments. AI systems increasingly influence how information is generated, curated, and consumed by voters, including through platforms and generative AI tools. While this creates opportunities, it also raises concerns around accuracy, bias, and the perceived authority of AI-generated information, with direct implications for voter trust and informed participation. At the same time, important opportunities are emerging. AI offers new tools to enhance election operations, including forecasting, logistics planning, and oversight functions, as well as to improve voter engagement and accessibility, particularly for historically underserved populations through multilingual and adaptive tools. Overall, these dynamics underscore that without targeted governance responses, AI may strain institutional capacity and public trust—yet, if properly governed, it can significantly strengthen electoral integrity, inclusiveness, and resilience.

The Global Dialogue on AI Governance can play a critical role in advancing international cooperation by serving as a structured, multi-stakeholder platform that bridges global principles with sector-specific realities and operational needs. First, the Dialogue can facilitate sustained engagement among diverse stakeholders, including public authorities, technical experts, civil society and private sector actors. In sectors such as elections—where responsibilities are fragmented across institutions and jurisdictions—this type of structured engagement is essential for building shared understanding and aligning approaches across actors who otherwise engage only reactively or infrequently. Second, it can promote cooperation by surfacing and addressing sector-specific governance gaps that are not visible through sector-neutral discussions. By enabling focused examination of how AI operates in specific institutional contexts, the Dialogue can support more coherent and responsive governance frameworks that reflect real-world operational constraints and societal impacts. Third, the Dialogue can play a key role in translating high-level global norms into practical, implementable guidance. International cooperation is often hindered by the gap between principles and practice; by supporting the development of shared approaches on issues such as oversight, auditability and procurement, the Dialogue can make cooperation more tangible and actionable across jurisdictions. Fourth, it can strengthen coordination across interconnected issue areas, recognizing that AI governance challenges frequently span multiple domains (e.g., institutional integrity, information environments, and market dynamics). By linking these areas, the Dialogue can help ensure that governance responses are coherent and mutually reinforcing. Finally, the Dialogue can contribute to international cooperation by grounding AI governance in shared societal outcomes, including public trust, institutional legitimacy, and inclusive participation. By centering these outcomes, it can help align diverse actors around common objectives, even in the absence of uniform regulatory frameworks. In sum, the AI Dialogue can advance cooperation by moving beyond abstract consensus toward practical, coordinated, and context-sensitive governance pathways.

The AI Dialogue can build on a growing ecosystem of multi-stakeholder, sector-focused initiatives that are already advancing practical approaches to AI governance, while adding significant value through coordination, scaling, and integration. One key initiative is the AI Advisory Group on Elections (AI AGE), convened by IFES. AI AGE brings together election officials and AI experts to close a critical gap between technical and institutional communities, with the goal of ensuring that AI development and governance safeguard electoral integrity and strengthen civic participation. It serves as a model for how sector-specific expertise can inform global AI norms, while also translating those norms into practical guidance for electoral stakeholders. Similarly, the Global Network Initiative (GNI) offers a well-established example of multi-stakeholder governance in the technology sector, bringing together companies, civil society, investors, and academics to advance freedom of expression and privacy online. Its principles and implementation guidelines provide a structured framework for operationalizing human rights commitments, fostering accountability through independent assessments, and enabling collective action across stakeholders The AI Dialogue can add value by connecting and scaling these types of efforts across sectors and geographies. In particular, it can: Bridge sector-specific initiatives (like AI AGE) with broader governance frameworks (like GNI) to ensure that real-world insights inform global standards; Facilitate cross-sector learning, helping replicate successful models of multi-stakeholder coordination and accountability in other domains; Strengthen coherence across fragmented initiatives, addressing gaps where governance remains siloed; and Translate shared principles into coordinated, practical approaches, particularly in high-stakes areas such as elections. The AI Dialogue's added value lies in linking existing expertise, fostering sustained collaboration, and ensuring that governance approaches are both globally aligned and operationally grounded.

Different stakeholders can contribute to the AI Dialogue by bringing complementary expertise, operational experience, and accountability mechanisms, while the Dialogue itself should be structured to enable sustained, practical, and context-sensitive engagement. First, public institutions, including regulators and sectoral authorities such as election management bodies (EMBs), can contribute by articulating real-world governance needs and constraints. Their participation ensures that global discussions are grounded in institutional realities. Second, technical experts and private sector actors can provide insight into AI system design, deployment, and limitations, particularly in areas such as auditability, explainability, and content governance. Given the reliance of public institutions on externally developed systems, their engagement is essential for addressing governance gaps related to vendor opacity and system complexity. Third, civil society and independent experts can play a critical role in advancing human rights, inclusion, and public accountability, ensuring that governance frameworks reflect diverse societal impacts, including those affecting marginalized communities. Fourth, multistakeholder initiatives—such as the AI Advisory Group on Elections (AI AGE) and broader coalitions like the Global Network Initiative—demonstrate the value of structured collaboration, shared learning, and practical guidance, offering models the Dialogue can build upon. In terms of format, the AI Dialogue would benefit from: - Issue-specific working groups, enabling deeper engagement on high-impact domains (e.g., elections); - Regular, sustained engagement mechanisms, rather than one-off convenings; - Outputs focused on implementation, such as guidance on procurement, oversight, and accountability; and - Cross-cluster coordination, to reflect how AI governance challenges span multiple thematic areas. In sum, the Dialogue should be designed not only as a forum for exchange, but as a practical, multi-level platform that translates global principles into actionable governance approaches across sectors and contexts.

Global discussions on AI governance continue to reflect imbalances in whose knowledge, experience, and priorities shape decision-making, leaving several critical voices underrepresented. Addressing these gaps is essential to ensuring that governance frameworks are both legitimate and effective. First, stakeholders from the global South remain systematically underrepresented, despite being significantly affected by AI's deployment. Research shows that AI governance agendas are still largely shaped by high-income, AI-intensive countries, while perspectives from low- and middle-income contexts are marginalized. This limits the ability to anticipate context-specific risks and opportunities. Greater inclusion requires expanding participation in working groups, advisory bodies, and global convenings, alongside investments in local expertise and decision-making power. Second, civil society organizations and grassroots communities, particularly those representing marginalized populations, are often insufficiently integrated into decision-making processes. While frequently consulted, their influence on outcomes remains unclear and constrained. Meaningful inclusion would involve earlier and more substantive participation, clearer roles in governance processes, and sustained funding to support engagement. Finally, sector-specific public institutions, including election management bodies and other frontline governance actors, are underrepresented in global AI discussions. These actors operate within specific legal and operational constraints and are directly responsible for implementing AI systems in high-stakes environments. Their inclusion is essential to ensure that governance frameworks are operationally feasible and aligned with institutional realities. To address these gaps, the AI Dialogue should prioritize equitable representation, structured participation mechanisms, and sustained engagement, ensuring that governance is informed not only by technical expertise but also by diverse societal perspectives and real-world experience.

First, the Dialogue should incorporate issue-specific working groups with rotating leadership, allowing stakeholders from different sectors—such as public institutions, civil society, and technical experts—to co-lead discussions. Focused groups (e.g., on elections) can surface context-specific risks and translate high-level principles into practical guidance, while leadership rotation helps diversify perspectives and ownership. Second, scenario-based simulations and use-case workshops could foster deeper engagement. For example, participants could work through real-world governance challenges—such as AI deployment in high-stakes settings—examining trade-offs related to transparency, accountability, and operational constraints. These formats enable participants to move from abstract debate to applied problem-solving, grounded in institutional realities. Third, the Dialogue could adopt "paired dialogue", where actors who rarely engage directly—such as regulators and technologists, or civil society and private sector representatives—are brought together in structured exchanges. This approach reflects the value of bridging silos, helping to build shared understanding across communities with different expertise and incentives. Fourth, co-creation labs or policy design sprints could be used to jointly develop outputs such as guidelines on procurement, oversight, or auditability. These collaborative sessions can accelerate progress toward actionable deliverables, rather than remaining at the level of general principles. Finally, the Dialogue should integrate hybrid participation models, combining in-person and virtual engagement with structured input channels for underrepresented stakeholders. This helps ensure sustained, global participation beyond formal convenings. Together, these formats would support a Dialogue that is interactive, inclusive, and outcomes-oriented, enabling stakeholders not only to exchange perspectives but to jointly develop practical governance approaches aligned with real-world needs.

The most effective policies, practices and approaches translate high-level principles into operational mechanisms and bridge technical and institutional expertise. From IFES' work, a key example is the forthcoming Strategic Playbook on Elections Administration in an AI Age, practice-oriented guidance that supports election management bodies (EMBs) to make strategic decisions on AI adoption and deployment. This is an effort of the multistakeholder AI Advisory Group on Elections (AI AGE). This approach emphasizes translating global principles-such as transparency, accountability, and human rights-into actionable tools, including guidance on procurement, auditability, and oversight. It also promotes multistakeholder engagement, bringing together public authorities, technologists, and civil society to address real-world governance challenges across the AI lifecycle. Effective approaches equip civil society, public sector actors and policymakers with tools to engage technologists and technology companies, including structured processes for dialogue, accountability, and negotiation. These approaches demonstrate how governance frameworks can move beyond principles to practical engagement strategies that shape corporate behavior and public policy. Another important model comes from multistakeholder coalitions such as the Global Network Initiative (GNI), which combines principles, implementation guidelines, and independent accountability mechanisms. GNI's framework illustrates how governance can be operationalized through ongoing assessment, shared learning, and collective policy engagement across companies, civil society, and governments. Across these examples, several effective practices emerge: Embedding governance in real-world institutional processes, such as procurement and oversight; Creating structured mechanisms for multistakeholder collaboration and accountability; Developing practical tools and guidance, not just high-level norms; and Ensuring continuous learning and adaptation, rather than one-off policy interventions. Together, these approaches demonstrate that effective AI governance requires integrated, participatory, and implementation-focused models that are responsive to both technical realities and societal needs.