February 21, 2025
Maritza Chan Valverde
Permanent Representative of Costa Rica to the United Nations
Héctor José Gómez Hernández
Permanent Representative of Spain to the United Nations
United Nations Secretariat
405 East 42nd Street
New York, NY 10017
Re: United Nations Consultation on the Independent International Scientific Panel on AI and the Global Dialogue on AI Governance
Dear Excellencies,
My name is Ryan Nabil, and I serve as the Director of Technology Policy at the National Taxpayers Union Foundation in Washington, DC, where my research focuses on US and global technology and AI policy issues.[1] I welcome the United Nations’ continued commitment to AI governance, and appreciate the opportunity to submit the following written evidence in response to the UN’s consultation on the Independent International Scientific Panel on AI and the Global Dialogue on AI Governance.
The attached document contains my responses to the consultation questions.
I hope these insights prove useful as the United Nations continues to develop its AI policy initiatives and mechanisms to contribute more effectively to international AI governance.
Best wishes,
Ryan Nabil
Director and Senior Fellow, Technology Policy
National Taxpayers Union Foundation
122 C Street NW
Washington, DC 20001
I. Independent International Scientific Panel on Artificial Intelligence
Q1. What should be the mandate of the multidisciplinary Independent International Scientific Panel on AI, to be established within the United Nations?
The mandate of the Independent International Scientific Panel on AI should be to provide scientific, evidence-based perspectives on the governance and applications of AI in different areas that fall within the UN’s remit. As a neutral and scientific body, the Panel can play an important role in providing perspectives on how Member States can best utilise AI and mitigate potential risks associated with different AI uses. Its insights can also be helpful in focusing and specifying the scope of the UN’s work in relation to specific areas of AI governance.
However, caution is warranted in defining the Panel’s mandate and functions. First, the UN leadership must ensure that the Panel does not duplicate the work of existing AI initiatives, thereby avoiding the misallocation of limited UN resources. Instead, the Panel and the Dialogue should focus on functional roles and subject areas underserved by existing AI initiatives where the UN has a comparative advantage.
Second, the Scientific Panel’s work must remain evidence-based, non-partisan, and neutral. Safeguards should be in place to prevent undue political influence, helping ensure the integrity and impartiality of its work.
Third, while scientific consensus can be valuable, the Panel should not be discouraged from addressing complex AI governance challenges on which consensus might be unlikely or difficult to achieve. In such cases, the Panel’s proceedings and publications should reflect divergent perspectives and ongoing scientific debates rather than attempting to present the majoritarian view.
Finally, the Panel should adopt an informative rather than prescriptive role in AI governance discussions. Its primary function should be to produce independent, scientifically rigorous insights that can inform evidence-based policy making. While policy recommendations from the Panel can be helpful, the Panel should not seek to dictate outcomes, but rather present policy options that Member States can consider. Maintaining this focus on scientific expertise, academic rigor, and regulatory restraint will be essential to the Panel’s long-term effectiveness and credibility as an impartial and trusted voice in global AI governance.
Q2. What should be the size, composition and governance structure of the Panel?
The United Nations must consider two competing concerns in determining the size and composition of the Panel. First, given the multifaceted nature of AI governance, no single panel is likely to possess expertise across all aspects of AI and its applications in varied contexts, from arms control to public-sector and educational use. Second, while there may be an inclination to create a large panel to ensure representation from different academic disciplines and regions, an overly large and centralised body could hamper its effectiveness. Therefore, any decision on the size and composition of the Panel must balance these competing priorities.
Two organisational solutions could effectively address these challenges. First, the Panel could establish separate working groups or work streams to conduct rigorous, domain-specific research and analysis in a more focused setting. These groups should not be permanent but formed for a limited period based on the Dialogue’s priorities at a given time.
Second, the United Nations could establish a broader decentralised network of academics and scientists with expertise across multiple disciplines. While these members would not form part of the core Scientific Panel, they could be consulted on an ad-hoc basis by UN agencies and departments working on AI-related topics. Scientists and scholars with specific expertise could also be invited to join specialised working groups under the Panel’s umbrella.
Ultimately, a multi-layered structure with distinct functions may be more effective than an unwieldy centralised body attempting to oversee all aspects of AI governance. Nevertheless, some degree of experimentation and regulatory flexibility may be necessary before the United Nations identifies an organisational structure best suited to its institutional strengths and policy objectives.
Q3. How should the nomination and selection process of the Panel be?
The nomination and selection processes of the High-Level Advisory Body on Artificial Intelligence provide a helpful starting point for the Scientific Panel. A key strength of that process has been the openness of its nomination process. Nominations for the Scientific Panel should similarly remain open to all actors, including self-nomination, with experts selected based on merit.
Without such an open process, there is a risk that only scientists with close government ties—who may not necessarily be leading experts in their respective fields—would be nominated and selected. A better approach would be to maintain an open nomination system while implementing more rigorous selection and evaluation criteria to ensure that experts are chosen on the basis of their intellectual merit, academic and publications record, research contributions, and the ability to advance the Panel’s work.
However, the United Nations should consider enhancing its nomination and recruitment efforts for the Panel. Anecdotal evidence appears to suggest that while nomination announcements for the High-Level Advisory Board reached many within AI policy and UN circles, many scientists and scholars in leading universities and research centres—whose work would have made valuable contributions to the UN’s work—were unaware of opportunities. Expanding the UN’s outreach to top universities, research and policy centres, and other institutions of excellence would help ensure a qualified and varied pool of candidates for the Scientific Panel.
Furthermore, the United Nations should prioritise recruiting experts with a wide range of disciplinary backgrounds—from mathematics and computer science to international relations and comparative law—to foster a more nuanced and multidisciplinary approach to AI governance. While regional representation concerns are important, such concerns should not take precedence over merit-based selection. Instead, the primary objective of the selection process should be to appoint talented and distinguished academics and scientists performing pioneering research in their respective disciplines from different regions, irrespective of their nationality or institutional affiliation.
II. Global Dialogue on AI Governance
Q5. What should be the mandate of the Global Dialogue on AI Governance, to be established within the United Nations?
Whilst there is a growing number of AI governance related initiatives worldwide, many of these initiatives focus on regional or politically aligned states where consensus facilitates collective decision-making. The United Nations can distinguish itself by serving as one of the only global fora where states with differing perspectives and varying economic and regulatory capabilities can debate and exchange ideas related to AI risks, opportunities, and international cooperation.
However, the United Nations should be wary of overreach, which could undermine its long-term credibility and effectiveness as a constructive player in the evolving AI governance landscape. More specifically, the United Nations should not view the adoption of legally binding instruments or one-size-fits-all regulatory frameworks as its mandate for the Dialogue. The UN lacks the legal mechanisms and enforcement capacity to impose such regulations, and many AI governance decisions are best left to Member States—who can tailor their AI policies to their respective political systems, economic conditions, and regulatory objectives. Pursuing an overly ambitious and rigid approach would also detract from the UN’s ability to make more constructive but limited contributions to AI governance within its institutional remit.
Instead, the Dialogue should be designed primarily as a platform where different Member States and non-state stakeholders can discuss AI-related challenges and opportunities in areas where the UN can play a constructive role. The Dialogue can also serve as a forum for generating ideas and exploring potential mechanisms for economic and regulatory cooperation at the bilateral, regional, or multilateral levels.
Whilst formal agreements on certain aspects of AI governance might ultimately happen outside the UN among a more limited set of like-minded countries, the Dialogue could remain valuable as a neutral platform for stakeholders and countries with heterogeneous and, at times, divergent perspectives on AI governance. Such engagement opportunities are particularly important for many smaller and medium-sized developing countries, which often lack well-developed AI infrastructure and regulatory resources and tend to be underrepresented in existing AI governance discussions. By complementing its informative and consensus-building role with regulatory capacity-building initiatives, the United Nations can contribute constructively to AI governance within the limits of its institutional mandate.
Q6. What types of outcomes should it achieve?
Instead of viewing the Dialogue as an end goal, it might be better viewed as a platform to i) debate and exchange ideas related to public policies on artificial intelligence and its applications in different contexts and ii) enable action within and beyond the UN institutional framework. For instance, the Dialogue could serve as a venue to discuss ways in which Member States could better utilise AI applications to improve national educational systems or public-sector efficiency, while simultaneously addressing potential AI-related risks. Additionally, the sidelines of such conferences could facilitate informal meetings to discuss mechanisms for greater economic and regulatory cooperation at the bilateral and regional levels.
While it is possible that the Dialogue may contribute to limited agreements in some cases, such outcomes should not be viewed as the end goal. For example, if the Dialogue were to focus on narrowly defined topics where there might be a realistic prospect of international consensus among a sufficiently large number of Member States, it could facilitate the adoption of regulatory guidelines or frameworks. However, given the rapid evolution of emerging technologies, the heterogeneous composition of the UN’s membership, and the unlikelihood of broad consensus on many aspects of AI governance, such formal agreements should not be the primary objective of the Dialogue. On important but potentially contentious issues where no immediate consensus exists, the Dialogue can serve as a forum for evidence-based discussions and constructive engagement between stakeholders with differing perspectives. Such exchanges can enhance the public understanding of the multi-faceted challenges of AI governance in different areas, which can inform evidence-based, nuanced policy making at the national level.
Finally, the Dialogue must avoid duplicating institutional functions already fulfilled by existing AI governance initiatives, particularly those led by the Organisation for Economic Cooperation and Development (OECD), the Global Partnership on AI (GPAI), and the Internet Governance Forum (IGF). Instead, the Dialogue should focus on complementing existing multilateral efforts. To that end, the United Nations would benefit from establishing coordinating mechanisms with international AI policy initiatives, either through the Dialogue itself or ideally through a limited AI Office housed within the General Secretariat.
Q7. How should Governments and all relevant stakeholders be involved?
The United Nations should structure stakeholder involvement across multiple levels to align with the nature and objectives of different discussions. While broader, high-level exchanges are useful for general consensus building, more specialised issues may require dedicated working groups and expert panels. With that in mind, three main levels of stakeholder engagement should be considered.
First, the main Dialogue should facilitate exchanges between government leaders and officials, representatives from intergovernmental organisations, and stakeholders from academia, civil society and the private sector—enabling meaningful engagement between stakeholders who might not otherwise have the opportunity to interact directly.
Second, the Dialogue would benefit from mechanisms to solicit stakeholders input, including contributions from scientists and experts unaffiliated with the UN, business sector representatives, policymakers, and researchers from different regions.
Finally, the Dialogue could incorporate specialised work streams, allowing for deeper engagement among a smaller set of stakeholders on complex issues that require more detailed deliberation beyond high-level debates and consensus-building efforts.
Q8. What should be the format of the Dialogue?
A multi-layered format would enable the UN to facilitate meaningful participation from diverse stakeholders and play a more effective role in AI governance discussions. The main Dialogue—organised on the margins of existing UN summits—should function as a broad intergovernmental and multistakeholder forum, bringing together representatives from governments, intergovernmental organisations, the private sector, and the scientific community to discuss both general and thematic AI governance issues. However, because AI governance encompasses a wide range of sectors and contexts, the Dialogue should prioritise a limited set of thematic issues at any given time, particularly in areas where the UN enjoys a comparative advantage over AI governance initiatives.
Additionally, the Dialogue could benefit from specialised working groups or focused discussions on thematic AI policy topics. However, these should complement rather than duplicate the work of the Scientific Panel, the Internet Governance Forum, and other international initiatives, ensuring more efficient use of limited UN resources.
Finally, the Dialogue should remain open to and actively solicit input from non-state actors, particularly scientific experts and private-sector stakeholders at the forefront of AI research and development. In this regard, non-state actor participation should follow the model of the Internet Governance Forum, where they can contribute substantively to discussions, rather than that of the UN General Assembly, where their role is more limited. Beyond formal conferences and working groups, the Dialogue should implement additional participatory mechanisms, such as requests for evidence and open consultations, to gather input from the scientific community, civil society, and private sector of AI governance issues currently under discussion.
III. Relationship between the Panel and the Dialogue
Q9. What should be the relationship between the Panel and the Dialogue?
The Dialogue and the Panel should be designed to complement, rather than duplicate, each other and existing AI policy initiatives. Their roles are distinct: the Panel should function as an independent academic body that provides evidence-based policy insights, while the Dialogue serves as a broader multistakeholder political forum for debate and policy discussions. The Panel’s research and analysis can inform and enhance the Dialogue’s deliberations without constraining them.
This complementary role can be particularly relevant to thematic AI governance issues. For instance, the Dialogue could organise a summit on building regulatory capacity-building in developing countries. A specialised working group within the Panel could contribute by conducting independent analyses of AI policy challenges facing states with limited regulatory capacity and resources. The Panel could also evaluate the effectiveness of various regulatory approaches, such as different types of regulatory sandboxes, in promoting regulatory understanding and capacity in emerging AI applications. These research findings could provide a structured basis for more informed discussions within the Dialogue.
Likewise, the Dialogue can serve as an informal forum to discuss and facilitate agreements on regulatory capacity-building and knowledge-sharing. For instance, two or more Member States could discuss creating reciprocal Al sandboxes to promote regulatory learning and explore potential avenues for deepened regulatory and economic cooperation. To that end, individual Member States or specialised working groups, housed within the Dialogue or the AI Office, can provide boilerplate templates for regulatory sandboxes that interested countries could tailor to their specific economic and regulatory conditions.
Given the multi-faceted nature of AI governance, different working groups or sub structures might be required to address specialised areas. In such cases, the mandates and objectives of the Panel and Dialogue should guide their composition. The Panel’s subgroups should primarily comprise academic and scientific experts from various disciplines, enabling rigorous, evidence-based analysis. In contrast, given its policy-focused mandate, the Dialogue’s working groups should prioritise a diverse mix of experts and non-experts, including legislators and private sector representatives. If such subgroups are formed as part of the Panel and Dialogue, it is essential to ensure that they do not duplicate but complement each other’s efforts, allowing the UN to maintain coherence and efficiency in its AI governance initiatives.
Q10. How can the Panel and Dialogue effectively draw on and leverage existing initiatives within the United Nations? How can the UN system best support the Panel and Dialogue in a coordinated manner?
This question raises broader issues concerning the institutional design and strategic direction of the UN’s AI policy initiatives. In the past, several national leaders and policymakers called for the creation of an UN agency for AI regulation. However, as this author has previously argued in the Yale Journal of International Affairs, a more effective approach would be to enable existing UN agencies and departments to develop and incorporate AI-relevant work within the context of their respective remits, rather than creating a new institution with an overly broad and poorly defined mandate of global AI regulation. Since then, the UN leadership appears to have shifted away from such proposals in favor of more targeted initiatives—representing a step in the right direction.
As the UN further develops AI-relevant initiatives, such as the Dialogue and the Panel, it would benefit from the establishment of a small AI Office with a limited mandate focusing on coordinating the UN’s AI policy initiatives, ideally housed within the UN Secretariat. While care should be taken to limit its mandate and size, such a unit could help prevent duplication of efforts and ensure that the Panel, Dialogue, and any related initiatives operate efficiently, optimising limited UN resources.
Additionally, establishing coordination mechanisms between the AI Office and external AI governance intiatives—such as the IGF, GPAI, and OECD—would be crucial in ensuring that the UN complement rather than duplicate existing AI governance work. The AI office could also serve as a liaison platform for various UN agencies and departments, helping them develop AI policy expertise and facilitating interdepartmental collaboration on AI policy-relevant areas.
Finally, whilst mechanisms for greater engagement with different UN actors would enhance the Panel and Dialogue’s ability to focus and calibrate their work to the UN’s policy priorities and institutional strengths, the UN leadership must ensure that such mechanisms do not compromise the Panel’s independence. Care must be taken to prevent undue political influence, safeguarding the impartiality and integrity of the Scientific Panel’s work.
1. The opinions expressed in this document are those of the author and do not necessarily reflect the positions of the National Taxpayers Union Foundation.