By Javier Surasky
This is the original post version
Introduction
Global AI governance requires more than political declarations or normative agreements: it needs a shared knowledge base for assessing systems, comparing risks, and defining common priorities. This is why the first meeting of the UN Independent International Scientific Panel on AI matters.
The Panel may become a key instrument for reducing regulatory fragmentation, building a common vocabulary, and giving the United Nations a stronger role in the emerging architecture of AI governance: As António Guterres noted at the opening of the Panel’s first meeting, this means establishing “technical baselines” that can enable a better-grounded international conversation (Guterres, 2026).
What the Panel is
The IISPAI
was formally established by General Assembly resolution
A/RES/79/325 of 26
August 2025, which sets out its terms of reference and the modalities for its
creation and operation, and establishes the Global Dialogue on AI Governance. This followed the proposal
outlined in the report Governing AI for Humanity, produced by the United Nations
High-level Advisory Body on Artificial Intelligence. Moreover, this entire process is part of the Global Digital Compact adopted at the Summit of the Future in 2024.
The
resolution instructed the Secretary-General to draw up a proposed list of 40
scientists to serve on this body. That list was submitted to the General
Assembly and adopted on 12 February 2026, with the appointment of the 40 recommended experts (A/80/619).
The
IISPAI’s mandate is to produce independent, evidence-based scientific assessments of the opportunities, risks, and impacts of AI in order to provide
input to the future Global Dialogue on AI Governance, a mission that seems intended to
replicate, in the AI field, the role played by the Intergovernmental
Panel on Climate Change.
IISPAI: first meeting
The Panel’s
first meeting took place recently, on 3 March of this year, and the information
available about it is limited. This forces us to distinguish between what has
been officially reported and what can be inferred from the Panel’s mandate and
the Secretary-General’s statements on the matter (Guterres, 2026; United Nations, 2026).
We know
that, operationally, at that meeting, the IISPAI members chose Yoshua Bengio
and Maria Ressa to facilitate the drafting of its first report.
The
selection of these two figures is the IISPAI's first clear message: Yoshua
Bengio is a central figure in the development of deep learning, a professor at
the Université de Montréal, and a recipient of the 2018 Turing Award. Maria
Ressa is a Filipino journalist and the 2021 Nobel Peace Prize winner for her
defense of freedom of expression. Choosing both of them brings together
technical authority and a trajectory linked to the defense of rights and public
responsibility.
What was discussed
The
information about the meeting's content is extremely limited. The most
important point in the official sources is agreement on the need to build
shared technical baselines to assess AI capabilities, performance, and risks,
facilitate international cooperation in this area, and reduce regulatory
fragmentation (Guterres, 2026).
The IISPAI
also appears to have discussed, as part of the mandate given to the body,
suggestions for testing AI systems and measuring their risks, using data and
comparable criteria that may be useful for international discussions.
The first
meeting, therefore, focused on organizing future work and giving structure to
the Panel, defining leadership and central lines of work without yet entering
into discussion of concrete measures to be proposed to the international
community. This does not make the meeting any less relevant. Rather, it
highlights the political importance of its work and makes clear that this is
not simply a matter of sound digital technique but, above all, of establishing
an AI knowledge infrastructure within the UN system.
This
unwritten positioning, as I see it, is critical because the IISPAI stands at
the beginning of a longer process, and that initial position gives it an
advantage when setting the agenda. The Panel will provide input to the Global Dialogue on AI Governance, which is already in its
consultation phase and will culminate in a meeting to be held on 6 and 7 July
2026 in Geneva, and another in May 2027 in New York, back-to-back with the AI for Good Global Summit 2026, which will take place in the same city from 7
to 10 July.
Seen from a
distance, this seems to indicate that the UN is moving toward the creation of
an expert-based mechanism to reduce information gaps in AI, one that can also
lay the foundations for a common vocabulary for international conversations on
the issue, while still far from any sign of a central regulatory authority. Put
differently, the IISPAI’s meta-mission is to generate cognitive rather than
normative legitimacy.
Conclusions
Despite
limited public disclosure, the IISPAI’s initial meeting appears to have
clarified how its members view their role within the emerging UN architecture
for AI governance: producing evidence, organizing knowledge, and helping
establish common criteria for assessing systems and risks. This is a necessary,
though not sufficient, condition for discussions on global AI governance to
avoid being trapped in vague declarations and disconnected responses.
For that to
be possible, the IISPAI must translate its scientific authority into effective
political influence. That, in my view, is its main challenge: to give substance
to the often-invoked interface between science and policy, and to do so for the
benefit of people, in line with the principles enshrined in the Charter of the
United Nations, which remain the best representation of the non-negotiable dream of a world of
peace, justice, and equity. For this to be possible, the Independent International Scientific
Panel on Artificial Intelligence must, from the outset of its work, move toward
proposing minimum common criteria for testing AI systems and measuring risks
that can also be applied by States with limited technical capacities. If the
IISPAI is to be a seed of global digital governance, it should ensure that its
contributions include the voices excluded by that ecosystem.
Update
This article discusses the first meeting of the UN Independent International Scientific Panel on AI as an early step in a longer institutional process. The Panel is expected to provide input to the Global Dialogue on AI Governance, with meetings scheduled in Geneva in July 2026 and New York in May 2027.
Future developments may clarify whether the Panel can move from technical assessment to real political influence within the emerging architecture of global AI governance.
To visit the Independent International Scientific Panel on Artificial Intelligence website, click here