On 26 October 2023, the UN Secretary-General António Guterres launched a High-Level Advisory Body on Artificial Intelligence (HLAB on AI), composed of 38 experts tasked with making preliminary recommendations on the international governance of AI. The HLAB on AI released its Interim Report ‘Governing AI for Humanity’ on 21 December 2023.
The Simon Institute for Longterm Governance has reviewed the Interim Report to present its suggestions for the next phase of work and how to focus the Final Report, so that it can be of greatest use to policymakers in driving forward multilateral AI governance.
Brief summary of the HLAB on AI’s Interim Report
Note: The below is a neutral summary, adopting the language of the report itself.
- Read our full summary and response to the Interim Report here -
In this report, the HLAB on AI succinctly summarizes AI’s main opportunities and enablers, alongside its major risks and challenges. Noting that AI’s development and rewards are currently unevenly distributed amongst countries, it emphasizes the urgent need for governance to ensure that we harness AI’s potential “in ways that leave no one behind”, but points out that there is currently an important governance deficit, especially at international level.
Although the HLAB on AI asserts that addressing this global governance deficit requires novel functions and institutional arrangements, it stops short of proposing an institutional model for AI governance. Instead, it sets out five guiding principles and seven broad functions that it deems necessary for any global AI governance framework, which will need to leverage existing, as well as new, institutions.
Guiding Principles:
- AI should be governed inclusively, by and for the benefit of all - governance mechanisms must promote both equitable participation in AI development, deployment and use, as well as in AI governance (through lowering barriers to participation).
- AI must be governed in the public interest - private incentives (e.g. of technology companies) cannot be allowed to dictate AI governance, and governance efforts must consider public policy goals related to equity, inclusion, sustainability, societal and individual well-being, competitive markets and healthy innovation systems.
- AI governance should be built in step with data governance and the promotion of data commons - regulatory and legal frameworks that protect privacy and security of personal data will be a critical complement to AI governance arrangements, and public data commons that are relevant to solving societal challenges should be developed.
- AI governance must be universal, networked and rooted in adaptive multi-stakeholder collaboration - universal buy-in by Member States is key for building an interoperable governance framework that should consist of a network of new and existing institutions, with structures to effectively engage the private sector, civil society and governments.
- AI governance should be anchored in the UN Charter, International Human Rights Law, and other agreed international commitments such as the Sustainable Development Goals - these offer a robust foundation and compass, and AI governance will require the strong involvement of all Member States to empower and resource an effective response.
Institutional Functions:
- Assess regularly the future directions and implications of AI
- Reinforce interoperability of governance efforts emerging around the world and their grounding in international norms through a Global AI Governance Framework endorsed in a universal setting (UN)
- Develop and harmonize standards, safety and risk management frameworks
- Facilitate development, deployment and use of AI for economic and societal benefit through international multi-stakeholder cooperation
- Promote international collaboration on talent development, access to compute infrastructure, building of diverse high-quality datasets, responsible sharing of open-source models, and AI-enabled public goods for the SDGs
- Monitor risks, report incidents, coordinate emergency response
- Compliance and accountability based on norms
In the next phase of its work, the HLAB on AI will consult with stakeholders to discuss the issues and questions identified in the Interim Report. More information is available online at: https://www.un.org/en/ai-advisory-body.
Our response
We recognize the comprehensive scope of the HLAB on AI’s work and the speed at which it was prepared, given the urgency of the AI governance challenge. We also appreciate how the report concisely covers all of the major opportunities and risks of AI. In particular, we commend its taking a practical and pragmatic ‘forms follows function’ approach of first identifying key governance functions before considering potential models for future AI institutions.
Leveraging our expertise on AI governance and our experience working with Member States, we focus our response on how the content of this Interim Report can be complemented or reframed to provide practical and clear direction for Member States as they negotiate the Global Digital Compact (GDC) in 2024.
- Read our full summary and response to the Interim Report here -
In summary, we propose that:
- AI is not a single technology; different governance approaches are required for simple models vs ‘frontier AI’ research and development, and the largest governance gap currently concerns frontier AI.
- To aid Member States in developing an international governance regime, the HLAB on AI could further elaborate upon its proposed (sub)functions, to clarify the role of States, the role of existing institutions, and to estimate associated costs.
- Civil society actors could help to fill the important gap in terms of providing capacity building for Global South countries, to promote their participation in AI governance discussions over the coming months.
- A recommendation from the HLAB on AI regarding key definitions and standards (such as those developed by the OECD or ISO) would help to ensure that Member States are on the same page, promoting fruitful discussions during the next phase of work.
- The HLAB on AI should privilege interactions with Member States (rather than other, diverse stakeholders) to maximize alignment with ongoing processes, notably the GDC negotiations.
In terms of the implications of this Interim Report for the GDC negotiations, we note that:
For the GDC process
- The importance of inclusive participation by Member States comes through very strongly throughout the HLAB on AI’s report. This importance applies equally to the GDC process, which will face similar challenges to those outlined for AI. Rapid capacity building support for Global South countries is needed.
For the GDC content
- The HLAB on AI’s report lays out the key opportunities/enablers and risks/challenges clearly and concisely, providing an excellent reference for the scope of the GDC on AI-related issues. This should improve further in the next phase of work with the development of the risk assessment framework and, hopefully, of more key definitions.
- However, the current report insufficiently distinguishes between different kinds of AI (frontier AI vs other kinds of AI) and their varying levels of opportunity and risk. This makes it more difficult to set governance priorities, to align national and international governance efforts, and to sufficiently address high consequence, global risks (even if these are of low probability). The GDC should thus recognize that AI is not a single technology, specifically addressing frontier AI governance.
- Regarding the institutional functions and subfunctions identified by the report as necessary for a global AI governance regime, once further elaborated, these could form the basis for the GDC’s AI related content. The GDC will effectively provide the mandates to reform existing institutions, to create new ones and to allocate functions to specific actors - outlining the functioning of an international regime complex for AI.
- Read our full summary and response to the Interim Report here -