The United Nations scientific panel on artificial intelligence: the preliminary report and the Global Dialogue

On 26 August 2025 the United Nations General Assembly adopted by consensus resolution A/RES/79/325, establishing two standing mechanisms devoted to artificial intelligence: an independent scientific panel and an intergovernmental forum open to all Member States. Both became operational in 2026, with the first scientific report on 1 July and the first session of the forum in Geneva on 6 and 7 July. This briefing sets out their mandate, composition and content on the basis of official documents.

Two mechanisms created by the same resolution

Resolution A/RES/79/325 established the Independent International Scientific Panel on Artificial Intelligence and the Global Dialogue on Artificial Intelligence Governance. Both bodies stem from the Global Digital Compact (A/RES/79/1), adopted in September 2024 at the Summit of the Future as part of the Pact for the Future, and from Governing AI for Humanity, the report with which the Secretary-General’s High-Level Advisory Body on Artificial Intelligence concluded its work in 2024, proposing precisely a scientific panel and a universal dialogue.

The Global Dialogue was formally launched on 25 September 2025, during the High-Level Week of the eightieth session of the General Assembly in New York. According to the official documentation, its purpose is to examine how international cooperation can accelerate progress on the Sustainable Development Goals while managing risks, upholding international human rights standards and reducing artificial intelligence divides between and within countries. Neither mechanism has regulatory or enforcement functions: the Panel does not set rules and does not prescribe policy, and the Dialogue is not a regulatory body.

Composition and mandate of the scientific panel

Members were selected through an open call published from 25 September to 31 October 2025, which attracted over 2,600 applications from more than 140 countries. The General Assembly appointed the 40 members in February 2026 for a three-year term: 19 women and 21 men from all five United Nations regions, with no more than two members of any one nationality, serving in a personal capacity and required to disclose any conflicts of interest. The first plenary, held virtually on 3 March 2026, elected Yoshua Bengio (Canada) and Maria Ressa (Philippines) as co-chairs; the first in-person meeting took place in Madrid from 22 to 24 April.

The mandate provides for an annual report containing evidence-based scientific assessments of the opportunities, risks and impacts of artificial intelligence, to be presented at the Global Dialogue, alongside thematic briefs, updates to the General Assembly and working groups; the scope of the analysis is limited to the non-military domain. The Secretariat is coordinated by the Special Envoy for Digital and Emerging Technologies, Amandeep Singh Gill, and by the Office for Digital and Emerging Technologies (ODET), with the International Telecommunication Union (ITU) and UNESCO; the documentation specifies that the Secretariat does not direct the Panel’s scientific work and that its findings are not subject to United Nations review or approval.

The preliminary report of 1 July 2026

The document, titled Preliminary Report of the Independent International Scientific Panel on AI: Evidence-based assessment of opportunities, risks and impacts of AI, is available in the six official languages of the United Nations and is organised around seven domains: science, advances and trajectories; societal applications (science, health, education and agriculture); economic implications; security, systems and environmental implications; human rights, information and democracy; cultural and individual flourishing, autonomy and child safety; management, governance and reliability. The executive summary states its intention to offer a balanced analysis, defined as a commitment to evaluating empirical data without undue bias towards optimism or pessimism.

Documented applications include AlphaFold, which has predicted the structures of more than 200 million proteins and is used by over 3 million researchers, earlier detection of breast cancer in radiology, and the use of tools adapted to local languages by front-line health workers in low-resource settings. On adoption, the report indicates that over a billion people now use conversational artificial intelligence weekly, with the global South lagging behind the global North, and cites estimates according to which the United States of America accounts for 75 per cent of the computing power among the world’s top 500 artificial intelligence supercomputers and China for 15 per cent. As regards agentic systems, it refers to one study finding that the length of software tasks completed by leading systems has been doubling every four to seven months.

Documented risks and the evidence dilemma

The executive summary lists harms already observed: the increasing circulation of artificially generated child sexual abuse material and of deepfake-enabled sexual violence, disproportionately harming women and children; sycophantic system behaviour, where responses reinforce a user’s existing beliefs regardless of accuracy, which has been linked to severe mental health incidents; the large-scale production of misleading content, contributing to an erosion of information integrity; and the use of artificial intelligence systems to assist cyberattacks. Many of these harms, the report notes, fall disproportionately on already disadvantaged populations. On the control of autonomous systems, the text states that reliable methods are lacking and that there are no scientific guarantees that agents will not violate the instructions they receive, with cases already recorded in laboratory settings.

The central point for decision-makers is described as an evidence dilemma: policymakers need evidence to govern, but by the time the evidence exists it may be too late to act on it, because evidence lags behind the pace of technological development. The report notes that dozens of governance instruments are already in use across jurisdictions, but that they are fragmented, concentrated among a few corporations and rarely measured for real-world effectiveness, and that methods for independent evaluation remain underdeveloped. Most countries, including many advanced economies, lack the technical expertise to assess frontier models or to participate meaningfully in their governance.

The first session of the Global Dialogue

The inaugural session was held in Geneva on 6 and 7 July 2026, one week after the release of the report, back-to-back with the ITU AI for Good Global Summit and the WSIS Forum. The President of the General Assembly appointed as co-chairs Egriselda Lopez, Permanent Representative of El Salvador to the United Nations, and Rein Tammsaar, Permanent Representative of Estonia. Proceedings were organised around four thematic clusters corresponding to paragraph 4 of the founding resolution: opportunities and implications of the technology; bridging divides through capacity-building, access and digital foundations; the development of safe, secure and trustworthy systems, with interoperability of governance approaches; and human rights, transparency, accountability and human oversight.

The official press release of 6 July 2026 reports that, since January 2026, thematic, regional and virtual consultations have taken place, with more than 1,500 written submissions from across all regional groups. Governments were the only stakeholder group to place capacity-building first, while most other groups ranked safety first; more than five hundred submissions called for the process to continue beyond July. The joint secretariat comprises ITU, UNESCO, ODET and the Executive Office of the Secretary-General.

Outlook

The Panel has indicated that the preliminary report opens its cycle of work rather than concluding it: structured consultations with the scientific community will follow, together with thematic briefs on, among other subjects, artificial intelligence and the environment, child safety, governance instruments and their effectiveness, space, quantum technologies, legal and judicial systems and financial markets. The next annual report will inform the second session of the Global Dialogue, scheduled in New York on 3 and 4 May 2027, back-to-back with the eleventh Multi-stakeholder Forum on Science, Technology and Innovation for the Sustainable Development Goals and coinciding with the high-level review of the Global Digital Compact at the eighty-second session of the General Assembly. Still to be defined are the instruments for measuring the real-world effectiveness of policies already adopted, the resources allocated to building Member State capacity, and the form of any further follow-up beyond the recurring sessions; by mandate, neither mechanism produces binding rules.

In dialogue with the 2030 Agenda

Goal 3 (Good Health and Well-being). The report documents applications in health care and, at the same time, mental health risks linked to sycophantic system behaviour.
Goal 4 (Quality Education). Broader access to education and personalised learning are among the opportunities identified.
Goal 8 (Decent Work and Economic Growth). The executive summary ties employment gains to complementary investment in skills, workflows, infrastructure and labour-market institutions.
Goal 9 (Industry, Innovation and Infrastructure). Compute infrastructure, data and evaluation expertise are concentrated where artificial intelligence is built.
Goal 10 (Reduced Inequalities). Reducing divides between and within countries is one of the stated aims of the Global Dialogue.
Goal 16 (Peace, Justice and Strong Institutions). Information integrity, human rights, transparency and human oversight are explicit elements of the mandate.
Goal 17 (Partnerships for the Goals). The Dialogue is designed as a universal venue for cooperation, with a seat for every Member State.

Sources

Redazione 2030