Opening Ceremony

The session formally opened the first United Nations Global Dialogue on AI Governance and framed it as the start of a longer multilateral process rather than a stand-alone conference . Ambassador Egriselda López of El Salvador presented the meeting as the launch of a broad and genuinely universal space where different experiences, capacities and actors could converge, building on a General Assembly mandate as well as years of prior innovation, knowledge generation and shared practice among states, international organisations, civil society, academia and technical communities . The co-chairs and secretariat were thanked for helping transform the member-state mandate into the dialogue now opening . López also stressed that Geneva should be seen not only as a place of arrival but as a point of departure for continued work . Speakers situated the Dialogue within the mandate created by the Pact for the Future and the Global Digital Compact, and several of them stressed that it should continue beyond this inaugural meeting .

López and Rein Tammsaar of Estonia set out the co-chairs’ basic framing of the meeting around inclusion, capacity and trust-building . López argued that the value of dialogue lies not merely in exchanging formal positions, but in sharing experiences, learning from one another and strengthening international cooperation in response to a technology that is evolving rapidly and therefore requires agile and inclusive responses . She also stressed that effective participation means more than physical presence: countries need skills, infrastructure, talent, financing, institutions, alliances and partnerships if they are to shape and benefit from AI on equal terms . Tammsaar reinforced this by noting that AI is already affecting every country regardless of technological development, so the conversation on governance must include every region, level of development and relevant stakeholder, with no one left behind . He described the co-chairs’ task as creating a universal platform where diverse perspectives, scientific knowledge, political judgment and technological deliberation could come together to build trust and common understanding .

This opening framing was tied to a second central idea: AI governance must balance opportunity with responsibility . López explicitly called for a balance between innovation and responsibility, opportunity and risk, international cooperation and national realities, and global decision-making and the concrete needs of individual countries . Tammsaar similarly argued that AI’s many opportunities can only be fully realised if systems are safe, rooted in international law and human rights, and shaped by people themselves rather than imposed upon them . Drawing on Estonia’s own experience, he said digital tools can significantly improve daily life when they are human-centred, trustworthy and supported by strong public institutions, while warning that technology on its own is never enough . He also clarified that the purpose of the next two days was not to erase differences, but to understand them better, identify areas of convergence, gather best practices and lay foundations for future cooperation .

António Guterres then shifted the discussion to urgency and warning . He said AI is advancing at “runaway speed” and is being deployed faster than anyone, including its builders, can fully keep up with, even though it can reshape economies, transform labour markets, sway elections and alter the balance of security . His most forceful formulation was that “an experiment is being run on our planet” and on societies “without a plan” and “without consent”, which he called both unsustainable and unacceptable . He framed the core question as whether humanity will shape AI’s transformation together or allow the technology to shape humanity instead, and said this Dialogue was meant to provide the beginning of an answer because, for the first time, every country had a seat at the table and a shared evidential basis .

A key part of Guterres’s speech was his reference to the first report of the Independent International Scientific Panel on Artificial Intelligence, which its co-chairs were due to present that morning, as the shared evidential basis for the discussion . He emphasised that the panel consisted of forty leading experts from all regions and disciplines serving in their personal capacities, independent of governments, companies and institutions . The creation of this panel was echoed later by Annalena Baerbock, who described it as a major milestone because it brought together expertise across sectors while also achieving gender and regional parity . Across the opening, speakers repeatedly linked political legitimacy to scientific legitimacy by stressing both broad participation and an independent evidence base .

Guterres structured the risks around three warnings from the science: speed, power and truth . On speed, he contrasted the internet’s fifteen years to reach one billion people with AI’s two years, and warned that AI systems are no longer merely passive tools but are increasingly writing code, acting online and making choices with diminishing human oversight . His point was that institutions were built to govern machines that follow commands, not machines that decide . On power, he warned that compute, data and talent are concentrated in only a handful of companies and countries, leaving most nations - especially many developing ones - without meaningful influence over decisions that will shape their futures . He expressed this as a structural inequality problem: if power imbalances are hardwired into technology, then inequality becomes part of the code itself . On truth, he warned that AI-enabled lies can now persuade as effectively as truths and that genuine evidence can be dismissed as fake, eroding trust and weakening the ability of societies to agree on reality or defend themselves . He sharpened this point through the phrase “vibe coding”, criticising the temptation to let AI systems operate because they seem to work without properly understanding or scrutinising them, and insisted that humanity cannot “vibe code” either truth or its own future .

Even while highlighting these dangers, Guterres also made a sustained case for AI’s developmental and humanitarian promise . He gave practical examples of AI helping a mother in a rural clinic receive a cancer diagnosis in time, a child continue learning with a tireless tutor, and a small farmer gain access to forecasts previously available only to the biggest agribusinesses . From these examples he drew a broader historical argument: expertise has long been scarce, expensive and concentrated, but AI could make it more widely available far faster than earlier transformational technologies such as penicillin or electricity . If used well and shared widely, he argued, AI could compress decades of development into years and become a “great equalizer” of the twenty-first century . This aligned with López’s opening claim that AI should contribute to development and deliver real benefits for all , and with later examples from Baerbock about disaster warning, agriculture, health and education .

Guterres then connected the Dialogue to a broader UN process . He recalled that the Pact for the Future and the Global Digital Compact had created the mandate for this work and said the opening of the Dialogue, the presentation of the scientific panel’s report and the discussions later that week at the AI for Good Summit should be seen as part of an integrated UN effort . He presented this as a way of bringing together political discussion, scientific evidence and practical initiatives within one wider governance process .

From there, Guterres moved to a governance agenda organised around four priorities: safety, red lines, capacity and transparency . On safety, he argued that alignment across countries on testing, risk measurement and responsibility is essential; otherwise, fragmented rules will raise costs, divide the world and still fail to protect people . He called for common baselines for frontier systems, shared methods for evaluating and verifying risks, and a common resolve that systems with global reach should meet standards worthy of global trust .

Within the safety agenda, Guterres gave special priority to children, arguing that no child should be treated as a guinea pig for unregulated AI . He contrasted the safeguards applied to medicine and toys with the lack of equivalent testing for AI systems already reaching children in their learning, friendships and private questions . He cited harms already visible: children deceived by machines posing as friends, children steered towards self-harm, and children abused through sexual images generated at the touch of a button . On that basis he called for an AI child safety pledge containing three rules: first, any system accessible to children should be proven safe through child-specific testing and independent oversight; second, companies should maintain zero tolerance for the generation of sexual images of children and should detect, report and remove such material; and third, when a child shows distress, the system must stop and connect them to real human support . This focus on children later found support in UNESCO’s announcement of a collective reflection on a new normative instrument to safeguard children and young people in the AI era .

On red lines, Guterres said human rights are not negotiable and that AI must never strip people of dignity or entrench discrimination . He insisted that in all high-stakes domains - justice, health care and policing - machines may inform decisions, but humans must remain the ones who decide and who answer for those decisions . This rights-based framing echoed Tammsaar’s insistence that AI must be grounded in international law and human rights , and later matched UNESCO’s presentation of its 2021 Recommendation on the Ethics of Artificial Intelligence as the first global standard aimed at aligning AI with human rights, sustainability and inclusion .

On capacity, Guterres highlighted the gap between massive private AI investment and negligible public investment for developing-country capacity . He warned that the digital divide could harden into an AI divide, which could then become a development gap, a security gap and a sovereignty gap . As a response, he noted that more than twenty member states had already nominated centres to a UN-supported global network for exchange and cooperation on AI capacity building . He also announced that he would submit recommendations to the General Assembly for a global fund for AI to build skills, data resources and affordable computing power “everywhere”, especially for developing countries . This directly reflected López’s earlier warning that meaningful participation requires infrastructure, financing and institutional capacity, not merely presence in a room , and Singh-Gill’s later remark that without capacity, dialogue becomes monologue and science remains abstract .

On transparency, Guterres argued that AI may feel intangible but its footprint is not . He pointed to the immense electricity and water demands of data centres, warning that by 2030 they could consume more electricity than all but five countries and enough water to meet the annual needs of all 1.3 billion people in sub-Saharan Africa . He added that communities hosting such infrastructure often bear these burdens without adequate information or benefit . He therefore referred to an AI Environmental Transparency Initiative calling on major AI companies to measure and publicly disclose the carbon, water and land footprint of their systems and to commit to powering every data centre with renewable energy by 2030 . This widened the AI governance conversation beyond safety and fairness to include sustainability and distributive justice.

Guterres also briefly but strongly addressed peace and security, even while noting that the Dialogue itself focused on civilian AI . He warned that the same models and chips used in civilian systems are also moving into the battlefield . His main concern was lethal autonomous weapon systems, which he deliberately labelled “killer robots” - machines selecting and engaging targets and taking lives without human control or judgment . He called such systems morally repugnant, politically unacceptable and something that must be banned by international law, insisting that some decisions must remain forever human and none more so than the taking of a human life .

In his conclusion, Guterres argued against the idea that governance and innovation are opposites . He said innovation needs guardrails, and that the technologies people trust most - in aviation, medicine and nuclear energy - are trusted because their makers are held to account . His summary formula was that if AI is to be powerful it must be governed, if it is to be trusted its builders must be accountable, if it is to be global it must be fair, and if it is to serve the future it must not consume the future . He then said that governments, companies, scientists and this dialogue itself all had urgent responsibilities, and warned that the present generation may be the last able to set the terms on which humanity and machines coexist . He closed by urging that this meeting be remembered as the moment governance began to catch up with technology and by looking ahead to reconvening in New York next year to assess whether AI is becoming safer, fairer, more accessible and more ethical .

Annalena Baerbock, President of the General Assembly, reinforced many of these themes while adding a strong defence of the UN’s institutional legitimacy and a frank recognition of policymakers’ knowledge limits . She said that what sounded like science fiction ten years earlier now already seemed outdated, and pointed to AI’s integration into software development and its growing ability to interact with the physical world through “world models” . Her central point was that previous technological revolutions unfolded over time, but in the case of AI “we do not have that luxury” . She contrasted this accelerated transformation with aviation, the steam engine and nuclear technology, all of which evolved more slowly or were deliberately slowed because of risk . She then asked how many policymakers truly understand the depth of changes associated with model architecture, autonomous agents or quantum computing, using that question to justify the need for broader and better-informed multilateral governance .

Baerbock argued that AI’s potential and peril are too significant and too far-reaching to be left to a few actors . She said that because AI reaches into economies, social systems, defence, peace and security, homes, food systems and even children’s bedrooms, it can only be meaningfully and safely managed together . She explicitly rebutted questions about whether the UN was acting within its mandate, arguing that the Charter itself provides the basis for international cooperation on a challenge like AI . She therefore answered potential scepticism by saying that AI needs a global platform and that this means it needs the United Nations .

Like other speakers, Baerbock stressed broad participation, but she framed it as an all-society responsibility . She said the world has learned that global problems require not just a few countries but all countries, and not only governments but also tech companies, civil society, societies and citizens . She also personalised this responsibility by saying AI governance needs participants not only as scientists, engineers and company owners but also as fathers and mothers because children are asking adults to act together . To underline this, she quoted a teenager speaking earlier at the General Assembly about hate travelling through screens and pixels and youth refusing to let platforms become battlefields .

Baerbock’s sharpest substantive addition concerned harmful uses of AI, particularly deepfakes and gendered abuse . She said adults, governments, policy-makers and businesses all have a responsibility to tackle sinister AI uses such as deepfakes, which disproportionately target women and girls . Citing figures that 99 per cent of deepfakes are sexual in nature and 96 per cent target women and girls, she argued that this is not accidental but systematic and represents a rollback of human rights and women’s rights . This brought a more explicit gender lens to the governance debate, complementing Guterres’s focus on children and broader information harms .

At the same time, Baerbock also emphasised AI’s positive potential for the Sustainable Development Goals . She cited examples from Bangladesh, where AI-powered systems provide families with instant flood and storm warnings, and from Kenya, where farmers receive detailed maps on where and when to plant seeds to maximise harvests . She then extended this to a broader UN imagination: UNICEF reaching children in crisis with personalised learning, and the World Health Organization detecting outbreaks before they become pandemics . Her aspirational conclusion was that this technological revolution should be one in which no country is left behind . She also described the scientific panel as a milestone because it brought together expertise across sectors with gender and regional parity .

The later institutional speakers focused more on implementation, coordination and continuity. Doreen Bogdan-Martin of the ITU welcomed participants to the first-ever UN Global Dialogue on AI Governance and drew a historical comparison with the internet . She noted that although the internet entered homes and businesses in the early 1990s, it took an entire decade before stakeholders gathered in Geneva at the World Summit on the Information Society to discuss how to make that technology people-centred, inclusive and development-oriented . By contrast, she said, within only three years of generative AI going mainstream, the UN had convened this week in Geneva to shape AI’s trajectory together . She situated the Dialogue within “Geneva Digital Week”, linking it to the scientific panel, the AI for Good Global Summit and the WSIS Forum, thereby presenting the week as a coordinated governance effort rather than a series of isolated events . She also thanked Switzerland as host country for its generous support . More broadly, she described the gathering as one in which the UN system, governments, companies, researchers, the technical community and civil society would discuss how to keep humanity at the core of perhaps the most consequential technology yet .

Bogdan-Martin further stressed dialogue itself as a governance method . Citing Pope Leo’s Magnifica Humanitas, she noted repeated emphasis on dialogue as a means of establishing basic agreements that can support a shared vision despite ideological and practical differences . She said that this is what participants were gathered to do: to build the common good through a shared vision, under the leadership of the Secretary-General and the co-chairs .

Khaled El-Enany of UNESCO focused on the gap between principles and implementation . He said that although more governments are adopting AI strategies and more institutions recognise the need for ethical, safe and trustworthy AI, a significant gap remains between ambition and implementation . He presented UNESCO’s role in two parts. First, UNESCO provides a normative foundation through its 2021 Recommendation on the Ethics of Artificial Intelligence, which he described as the world’s first global standard in the field and a framework for aligning AI with human rights, sustainability and inclusion . Second, UNESCO works with more than eighty countries to strengthen legal frameworks, institutional capacities, governance arrangements, skills and accountability mechanisms so that principles are translated into practice . He also noted investment in people, stating that more than 50,000 civil servants and judicial actors across 192 countries had benefited from UNESCO-supported training programmes and tools . Finally, echoing the child protection concerns raised earlier, he said UNESCO was launching a collective reflection on a new global normative instrument to safeguard children and young people in the age of AI and digital technologies . His concluding message was that AI governance is a shared responsibility requiring a one-UN approach and partnerships across governments, business, academia and civil society .

Amandeep Singh-Gill, Under-Secretary-General and Special Envoy for Digital and Emerging Technologies, underlined both the scale of participation and the need for continuity . He noted that more than 4,000 people had registered to be in the room and that over 170 countries were represented, joined by scientists, entrepreneurs, civil society, international organisations and voices from around the world . He specifically noted the presence of the scientific panel’s co-chairs, Yoshua Bengio and Maria Ressa, alongside these other participants . He used this to make a conceptual point: inclusion is not a one-off act or a box-ticking exercise . Instead, he said, inclusion in AI governance must continue to be strengthened because AI is too consequential to be shaped by a few . His most pointed formulation was that the conversation must be global, inclusive and grounded in evidence, but also backed by capacity, “without which dialogues are monologues and science is just abstract” . He then situated the meeting in a decade-long UN trajectory involving the High-Level Panel on Digital Cooperation, the Roadmap for Digital Cooperation, the Global Digital Compact, the High-Level Advisory Body on AI and the work on modalities for these mechanisms . Finally, he stressed that dialogue itself cannot be a one-shot exercise; it must continue iteratively to connect norms with practice, and said the second round in New York would likely be held back-to-back with the STI Forum, possibly on 3 and 4 May, with his office preparing to coordinate the next cycle .

Taken together, the opening session showed broad alignment on the overall architecture of AI governance, even if not yet on detailed institutional design . Speakers converged on the need for governance that is inclusive, UN-linked, evidence-based and multistakeholder; they also agreed that AI’s opportunities in development, education, health and agriculture are real, but that those benefits must be matched by safeguards around safety, human rights, accountability, child protection, truth, environmental sustainability and peace . Differences of emphasis were still visible. Guterres focused most strongly on urgent guardrails, including a child safety pledge, common safety baselines, environmental disclosure and a legal ban on autonomous weapons . López, Tammsaar, Bogdan-Martin, El-Enany and Singh-Gill placed relatively more emphasis on balance, dialogue, implementation support, capacity-building and iterative follow-up .

Tammsaar closed the opening by summarising the speakers’ contributions as having highlighted both AI’s immense opportunities and the need for stronger international cooperation to ensure those technologies contribute to sustainable development, inclusion and shared prosperity . He then formally moved the meeting into the presentation of the preliminary report of the multidisciplinary, independent scientific panel on artificial intelligence . The opening therefore established the Dialogue as the beginning of a continuing UN process intended to connect scientific evidence, political inclusion and institutional coordination to practical future action .