São Paulo Multistakeholder Guidelines: A breath of fresh air into digital governance debates

On April 29 and 30 April, NETMundial+10 met in São Paulo on its 10th anniversary. Back in 2014, NETMundial gathered at the critical moment of ICANN’s transition from US government guardianship to multistakeholder one, which happened in September 2015. Somehow, there is a temporal parallel. This year, NETMundial met as diplomats, tech people, and civil society have been debating the future of AI and digital governance in the context of the Global Digital Compact and the 20th anniversary of the World Summit of Information Summit (WSIS+20).

The São Paulo Multistakeholder Guidelines, adopted last week, provide a breath of fresh air in increasingly confusing debates on AI and digital governance. The guidelines offer a pragmatic path forward by building on a solid experience of what works and does not in the field of digital governance. A few points in the Guidelines are particularly relevant.

You can follow NETMundial+10 outcomes in the wider context of digital governance processes: Global Digital Compact and WSIS+20.

Anthropomorphic AI caveat: The avatar in this video is developed by AI to mimic human expressions. Avatar does not have free will or consciousness.

1. Beyond the false dichotomy of multilateral vs multistakeholder approaches

For too long, discussions around internet governance have been mired in a false dichotomy: multilateral versus multistakeholder approaches. The São Paulo Statement clarifies this misunderstanding by illustrating practical interplays between multilateral and multistakeholder processes. The bridge is created by using the word ‘respectively’ to describe the specific roles and responsibilities of governments, civil society, businesses, and other stakeholders. The ‘Respective Formula’ builds on the provision from the WSIS Tunis Agenda and goes a step further by providing concrete suggestions for fostering collaboration among stakeholders. To be precise, this interplay among various actors will be critical for future digital negotiations at the UN and in other digital spaces.

2. Avoiding the risk of ‘fake multilateralism’

The first out of 13 guidelines from São Paulo addresses the critical risk of “fake multistakeholderism”, triggered by power imbalances.  When trillion-dollar companies and marginalised communities are placed at the same discussion table without addressing the imbalance between them, the results of deliberations can be superficial at best and manipulative at worst. The guidelines propose steps to empower weaker stakeholders, ensuring their voices are heard and respected, moving towards a genuinely inclusive and effective multistakeholder model.

3. Tangible outcomes of digital processes

The São Paulo Guidelines also advocate for more tangible outcomes from the Internet Governance Forum (IGF) and other policy deliberations. For example, the IGF can issue policy recommendations, where appropriate, as was stipulated in 2005 by the Tunis Agenda of the World Summit on Information Society.

For a long time, the IGF has shied away from producing policy recommendations due to concerns that negotiation dynamics could stifle the open exchanges characterising the IGF deliberations. This risk can be addressed by new AI solutions. Built around Large Language Models, AI can streamline the drafting process by ensuring that inputs and submissions accurately reflect or align with the final text of the adopted recommendation. Each participant in the drafting process can follow the fate of their submissions, similar to tracing flights online. This would substantively increase ownership and trust in global policymaking. 

AI can also help in determining whether the negotiated text aligns with broader contextual values, ranging from the UN charter and human rights declaration to AI and digital governance principles. While negotiations will still be a matter of persuasion and trade-off, AI can make them transparent, informed, and inclusive. 

Back in 2006, the IGF was among the first UN organisations to use online participation, which is now part of the regular modus operandi in the UN system and diplomacy in general. Similarly, at the current pivotal moment, the IGF and digital governance communities could make another decisive innovation by using AI to negotiate recommendations through an open-source, transparent, and traceable AI platform.

4. Avoid governance inflation and duplication

NETMundial+10 addressed the growing risk of governance inflation and duplication. Last year alone saw a proliferation of AI governance initiatives, which were popping up almost weekly as fear-mongering narratives prevailed after the introduction of ChatGPT. Proposals were made to govern AI analogously to nuclear energy (International Atomic Energy Agency – IAEA), climate change (International Panel on Climate Change – IPCC), or science (CERN). Many tech developers and scientists signed numerous letters advocating for AI development, or arguing that it should be paused or stopped.

Fortunately, we have moved past this frantic reaction, allowing us to approach the future of AI with the calmness necessary for designing laws and policies. However, one consequence of the 2023 AI governance tsunami will have a longer impact, as tens of initiatives, proposals, and organisations  were launched. These have gained a life of their own, with many people and organisations investing themselves in such initiatives, and many are supported by considerable funding and the narrow business interests of tech companies.

The creation of a new policy ‘industry’ leads to the proliferation of over 1000 initiatives and proposals on AI and ethics. As with every inflation, the inflation of governance spaces would reduce their value and relevance.  In AI and digital governance, more is not better. For example, this inflation strains the already limited human and organisational capacities of actors from small and developing countries, thereby affecting inclusivity. The increasing number of initiatives, bodies, working groups, and other forms of AI and digital governance will further marginalise already marginalised actors.

Against this backdrop, it is unsurprising that the NETMundial+10 statement mentioned word duplication 7 times and called for action to stop and control this governance inflation.

In the practical way of avoiding duplication, one could use some sort of  ‘Bauhaus test‘ by checking if ‘form follows functions’: whether the new bodies and organisation are needed to, for example, perform functions of ethical AI developments, avoid tech monopolies, or whatever else society considers a priority function.

In this debate of what works and does not work in digital governance, one should not confuse a lack of political will with a lack of governance bodies. New initiatives and bodies cannot replenish the lack of political will for AI and digital solutions.

---

The NETMundial+10 provides timely reality checks and concrete proposals ahead of accelerating digital negotiations around the Global Digital Compact and WSIS+20. It will help reduce confusion and increase clarity in digital governance, which will be critical in striking the right trade-off for the future of AI and digital developments.