By Javier Surasky
.png)
Epistemic community rather than global government
The United
Nations is often the reference point in debates over the creation of global
governance for artificial intelligence: if AI is, by nature, transnational, the
logical step would seem to be to seek a common regime by drawing on the
structure of the international system’s most universal organization.
But, as we
build our AI power map, that starting point seems, at the very least,
misleading and somewhat naive.
The UN does
not control the material conditions of artificial intelligence: it does not
produce frontier models, semiconductors or chips; it lacks global computing
infrastructure and a digital market of its own, and it has no real capacity to
discipline the companies, laboratories or States leading the technological
race.
Its place
on the AI power map, therefore, does not come from the ability to impose rules
or create incentives for innovation, but from something more limited and, at
the same time, highly relevant in political terms: creating a global epistemic
community around artificial intelligence, from which a “common language” can
emerge that enables agreement-building and, eventually, some model of
governance.
Today, we
see the UN trying to bring together the different parts of its highly complex
system, governments, experts, the private sector, academia, and civil society around
that effort to construct a shared scientific and narrative framework on issues
such as risks, opportunities, rights, development, security, and
responsibilities in the field of AI. Rather than seeking to give rise to a
strong global regime, it is trying to create the minimum cognitive, political,
and institutional conditions for such a regime to be discussed.
This view
is reinforced when we see that the central decisions on AI are being made in
other spaces, many of them “clubs of States” (such as the G7, the OECD or the
BRICS), and in large private companies in the sector. What the UN does have is
the possibility of granting global legitimacy to those conversations, or
withholding it, and of linking them to its key issues: sustainable development,
human rights, peace and security. Nevertheless, the UN cannot turn its role
into effective authority over the global technological ecosystem.
That is
where the UN’s central tension lies on the AI power map.
What is new
is that the United Nations is working to institutionalize that epistemic
function with the support of some of its members, as we see in the adoption of
the Global Digital Compact, within the framework of the 2024 Summit of the
Future, which incorporated commitments on AI governance and created two
specific mechanisms: an Independent International Scientific Panel on AI and a
Global Dialogue on AI Governance, both aimed at developing a minimum
architecture capable of producing shared knowledge, organizing evidence and
sustaining institutionalized multilateral deliberation.
Even before
the Summit of the Future, Secretary-General António Guterres had created the
High-level Advisory Body on Artificial Intelligence, a multistakeholder
advisory body that served as an agenda-setting space and produced the report Governing
AI for Humanity, published in September 2024, recommending the creation of
an international scientific panel, a global dialogue on governance, a
capacity-building network, a global fund for AI, a global data framework, a
standards exchange and a small AI office within the United Nations.
The General
Assembly took part in that agenda and transformed it into mechanisms with
intergovernmental backing through resolution 79/325, of August 26, 2025, which established
the Independent International Scientific Panel on AI and the Global Dialogue on
AI Governance committed to in the Global Digital Compact.
The first
is the UN’s current group of independent experts on AI, made up of 40 members
recommended by the Secretary-General and appointed by the General Assembly in
February 2026, with a three-year mandate and the basic mission of producing
independent, multidisciplinary and evidence-based scientific assessments of the
opportunities, risks and impacts of AI.
The scope
of that function must be approached carefully, because the Panel has absolutely
no normative capacity, which leaves its possible impact tied to the real
receptiveness it receives from States, and to the attention paid to it by major
private providers when defining policies on financing, capacity-building,
standards, security, and development.
The risk of
“Responsible AI-washing” is latent.
With the
establishment of that Panel, the UN manages to intervene in the dispute over
legitimate knowledge about artificial intelligence, but within an ecosystem
where strategic information is concentrated in private laboratories, national
agencies, cloud companies, and research centers in technologically advanced
countries. Once again, its strongest card is its ability to confer legitimacy.
The Global
Dialogue on AI Governance has a different and complementary function: as a
meeting platform for governments, the private sector, academia, the technical
community, and civil society, it offers a political space to debate and put
into practice the inputs the Panel may provide.
Here we see
the second key element of the UN as an actor in the AI ecosystem: being a pole
in the dispute over the language used to present the digital race, setting
elements such as development, the digital divide, human rights, environmental
sustainability, and inclusion, among others, against innovation, productivity
or national security.
Along the
same lines, the UN has managed to produce some incipient broad consensuses, as
shown by the adoption, in March 2024, of a General Assembly resolution 78/265,
on safe, secure and trustworthy AI systems for sustainable development, by
consensus, making it one of the first broad political agreements on AI at the
universal level.
In December
2024, the General Assembly also adopted resolution 79/239 on artificial
intelligence in the military domain and its implications for international
peace and security, which took the debate beyond lethal autonomous weapons
systems and projected it onto the military use of AI, although this time a vote
was required, with 159 votes in favor, five abstentions (Belarus, Ethiopia,
Iran, Nicaragua and Saudi Arabia) and two votes against (DPR Korea and Russia).
Russia’s
negative vote is especially relevant and should be added to the digital and
geopolitical competition between China and the United States, since the
Security Council can hardly build rules on an issue where its permanent members
have divergent technological, military and political interests.
The UN’s
involvement goes beyond those bodies: UNESCO, part of the “UN system,” adopted
the Recommendation on the Ethics of Artificial Intelligence in 2021, the first
global standard on AI ethics, applicable to its 139 member states. Once again,
it is not a document that generates legal obligations, but merely political
ones, although it may begin to be regarded as soft law.
Besides, the
WHO has worked on ethics and governance of AI in health, the ILO in the field
of work, UNICEF on its impacts on children, and UN-Women on gender inequalities
in STEM and on models biased on gender grounds, and the list continues, but we
want to highlight the International Telecommunication Union (ITU), whose AI for
Good platform has become an internationally visible space for connecting AI,
standards, development and sectoral applications.
Here a
second role appears, which is also an internal challenge for the UN: generating
coherence and synergies within its system of entities to move from the
epistemological field to the creation of technical communities, applied
projects and cooperation among agencies. T
his will be
difficult, since the UN lacks a common AI policy that extends across its
system; instead, each entity develops its own. More seriously still, it is not
clear that the General Assembly and the Secretary-General are following a
common policy.
Nor does
the UN have a global fund of sufficient scale to finance its positions, leaving
it at a disadvantage compared to other actors capable of offering stronger
incentives, including financial ones. Without its own resources to invest in
AI, the risk is that the UN will end up “managing” inequalities rather than
reducing them.
Insisting
on the UN’s role in creating global AI governance as the central axis means
pushing it to produce declarations about a technology whose real power is
decided elsewhere. It would be, we believe, more intelligent to understand its
role as the construction of a global epistemic community: organizing
independent scientific evidence, mapping capacity gaps, connecting technical
standards with human rights, opening channels for the Global South,
coordinating agencies, pushing for financing for public digital infrastructure
and sustaining debates on military AI that are not monopolized by the powers,
without renouncing the idea of global AI governance but being strategic and
measuring objectives and capacities.
If the UN
manages to turn legitimacy into shared knowledge, coordination, evidence,
financing and interoperable standards, it will be able to influence global AI
governance from the place where it can be most effective: the construction of a
global epistemic community.
In a
technological ecosystem that advances at industrial speed while the
multilateral architecture moves at diplomatic tempos, the UN is too weak to
govern artificial intelligence, but also too important to ignore.
Basic facts
- The UN’s main power in AI lies in multilateral legitimacy, universal convening power, institutional coordination, the building of political consensus and, more recently, the construction of a global epistemic community.
- The UN Secretary-General created the High-level Advisory Body on Artificial Intelligence in 2023, a multistakeholder advisory body that produced the final report Governing AI for Humanity in September 2024, used as an input at the Summit of the Future, where the Global Digital Compact was adopted, taking from the report the recommendation to establish a minimum architecture for international cooperation on AI. As a result, the General Assembly established, through Resolution 79/325 of August 2025, the Independent International Scientific Panel on AI and the Global Dialogue on AI Governance.
- The Independent International Scientific Panel on AI is a group of independent UN experts on AI, made up of 40 members recommended by the Secretary-General and appointed by the General Assembly in February 2026, which must produce independent and multidisciplinary scientific assessments of the opportunities, risks and impacts of AI.
- The Global Dialogue on AI Governance functions as the Panel’s complementary political space: it seeks to bring together States and other actors to discuss AI governance, interoperability, capacity gaps, standards, human rights, development and security.
- In March 2024, the General Assembly adopted a resolution by consensus on safe, secure, and trustworthy AI systems for sustainable development, but in December of that same year, it had to put a resolution on artificial intelligence in the military domain and its implications for international peace and security to a vote.
- Since 2021, UNESCO has had a Recommendation on the Ethics of Artificial Intelligence, the first global standard on AI ethics.
- The ITU organizes AI for Good, the main platform of the United Nations system dedicated to applications, standards, capacities, and cooperation on AI for development objectives.
- The United Nations’ main strength lies in its universal legitimacy and its capacity to link AI with sustainable development, human rights, peace and security. Its main vulnerability is the lack of coercive power, insufficient financial resources, technological infrastructure, and capacity to discipline leading States and companies in the technology sector.