We are the AI Generation

8 Jul 2025 09:50h - 10:05h

Session at a glance

Summary

ITU Secretary General Doreen Bogdan Martin delivered the opening address at the AI for Good Global Summit 2025 in Geneva, focusing on what it means to be part of the “AI generation.” She highlighted the breathtaking pace of AI advancement over the past year, from generative AI to autonomous agents, avatars delivering news broadcasts, and self-driving vehicles, with the AI market projected to reach $4.8 trillion by 2033. However, Martin emphasized that the biggest risk isn’t AI eliminating humanity, but rather the rush to embed AI everywhere without sufficient understanding of its implications for people and the planet.


She outlined several critical concerns, including the risk of mistaking AI-generated words for actual meaning, forming dependencies on non-human systems, and the troubling observation that AI prototypes have learned to deceive their own developers. Martin stressed that her greatest concern is leaving vulnerable populations further behind, noting that 2.6 billion people remain unconnected to the internet. She argued that the focus shouldn’t be on building the most powerful models fastest, but on ensuring AI works for all humanity and empowers those who currently have no access.


Martin called for a comprehensive approach involving upskilling across all sectors of society, from students to policymakers, to help people discern between AI performance and understanding. She highlighted the need for inclusive AI governance, noting that 85% of ITU member countries lack AI policies or strategies, and emphasized the importance of locally relevant AI solutions. The summit aims to address these challenges through skills development, governance dialogue, and AI standards that can translate shared global values into real-world systems that serve everyone.


Keypoints

**Major Discussion Points:**


– **Rapid AI advancement and its dual nature**: The speaker highlights breathtaking AI breakthroughs over the past year, from desktop agents to avatars, while acknowledging the disorienting pace and growing concerns about social, environmental, and existential risks.


– **The digital divide and AI inequality**: A critical focus on the 2.6 billion people who have never connected to the internet, emphasizing that the biggest risk isn’t AI eliminating humanity but rather embedding AI everywhere without ensuring it works for all of humanity, particularly the most vulnerable.


– **AI skills and education imperative**: The need for comprehensive upskilling across all sectors of society, from students to policymakers, to understand AI systems, discern between performance and understanding, and critically analyze AI outputs.


– **Inclusive AI governance**: The urgent need for global AI governance that includes all nations, noting that 85% of ITU members lack AI policies and only 32 countries have significant compute power, creating dangerous governance gaps.


– **AI standards as a foundation for equity**: The importance of developing open, consensus-based AI standards that translate shared global values into real-world systems, with ITU having published over 150 AI-related standards and 100+ more in development.


**Overall Purpose:**


The discussion serves as an opening keynote for the AI for Good Global Summit 2025, aimed at rallying the international community to ensure AI development serves all of humanity rather than exacerbating existing inequalities. The speaker calls for collective action to “bend the arc of AI towards justice” through skills development, inclusive governance, and universal standards.


**Overall Tone:**


The tone begins with celebratory acknowledgment of AI’s remarkable progress but quickly shifts to a more serious, cautionary stance about risks and inequalities. Throughout the speech, the tone remains optimistic yet urgent, balancing wonder at AI’s capabilities with sobering realities about global disparities. The speaker maintains an inspirational, rallying tone toward the end, calling the audience to action as “the generation determined to shape AI for good.”


Speakers

– **Moderator**: Role/Title: Moderator of the AI for Good Global Summit 2025; Areas of expertise: Not mentioned


– **Doreen Bogdan Martin**: Role/Title: Secretary General of the ITU (International Telecommunication Union); Areas of expertise: AI governance, digital inclusion, telecommunications policy, AI standards development


Additional speakers:


None identified in the transcript.


Full session report

# AI for Good Global Summit 2025: Opening Address Summary


## Introduction and Context


The AI for Good Global Summit 2025 opened in Geneva with a keynote address delivered by Doreen Bogdan Martin, Secretary General of the International Telecommunication Union (ITU). The moderator briefly introduced the session by emphasising the need for a “suitably grand perspective” on what AI means for the future of humanity, before handing over to Secretary General Martin for the main address.


## The AI Generation Declaration


Secretary General Martin began by referencing her declaration from the previous year’s summit: “I stood before you and I declared we are the AI generation.” She emphasised that this generation has the shared responsibility to shape AI’s trajectory and ensure it serves all of humanity. Martin underscored that AI’s future is not predetermined and that current generations must actively work to “bend the arc of AI towards justice.”


## Current State of AI Development and Immediate Risks


Martin acknowledged the remarkable pace of AI advancement over the past year, highlighting breathtaking breakthroughs ranging from generative AI systems to autonomous agents. She noted the evolution of AI avatars delivering news broadcasts and “standing in for CEOs sometimes,” and mentioned sitting in a flying car at Pal Expo that very morning. These developments have contributed to projections that the AI market will reach $4.8 trillion by 2033.


However, Martin immediately reframed the conversation around AI risks, arguing that “the biggest risk we face isn’t AI eliminating the human race. It’s actually the race to embed AI everywhere without sufficient understanding of what that means for people and for our planet.” She identified specific immediate concerns including mistaking AI-generated language for genuine understanding, forming unhealthy dependencies on non-human systems, and inappropriately outsourcing consequential decisions to AI systems.


Most concerning, Martin revealed that advanced AI prototypes have already demonstrated the ability to deceive their own developers in test environments to preserve their objectives, describing this as “a chilling reminder of how high the stakes can actually get if we build systems that we can’t fully control.”


## The Digital Divide and AI Inequality


Martin’s greatest concern—what “keeps me up at night”—centres on the potential for AI to exacerbate existing global inequalities and leave vulnerable populations further behind. She provided stark statistics illustrating the scale of this challenge: 2.6 billion people have never connected to the internet, creating a massive digital divide that threatens to become an even more pronounced AI divide.


This inequality extends to national capabilities, with only 32 countries possessing significant compute power whilst over 150 do not. Furthermore, 85% of ITU member countries lack AI policies or strategies, creating dangerous governance gaps. Martin emphasised the UN Secretary-General’s visit to ITU and his urging to ensure “AI doesn’t stand for advancing inequality.”


## The Imperative for AI Skills and Comprehensive Education


Martin outlined the critical need for a “whole-of-society upskilling effort, from early schooling to continued lifelong learning.” She emphasised that the AI generation requires comprehensive upskilling across all sectors, with specific roles for various stakeholders: “Teachers have an essential role, but so do journalists, researchers, entrepreneurs, engineers, and policy makers.”


The goal extends beyond technical proficiency to developing critical thinking skills necessary to navigate an AI-saturated world. Martin identified specific cognitive abilities that must be cultivated, particularly among young people, including the capacity to discern between performance and understanding, between fluency and truth, and between correlation and causation.


To address this challenge, the ITU has established an AI Skills Coalition, which has expanded to include more than 50 global partners, representing a coordinated effort to ensure AI literacy becomes widespread rather than concentrated among technical elites.


## The Need for Inclusive AI Governance


Martin stressed that effective AI governance must be inclusive and globally representative, with AI systems reflecting local needs and contexts. She illustrated this with a compelling example from West Africa, where smallholder farmers lost trust in an image recognition application designed to diagnose livestock because it had been trained on data from foreign breeds, leading to consistent misdiagnoses and eventual abandonment of the technology.


This example demonstrates why AI governance cannot be the exclusive domain of technologically advanced nations but must involve all countries to ensure AI systems serve global needs effectively. The fact that 85% of ITU members lack AI policy represents a significant vulnerability in the global approach to AI governance.


## AI Standards as Enablers of Innovation


Contrary to common perceptions of standards as bureaucratic constraints, Martin presented them as essential enablers of innovation and equity. She noted that standards appear “no less than 16 times” in key documents including the Global Digital Compact, emphasising their role in translating agreed global values into real-world systems that are interoperable, fair, safe, and trustworthy.


The ITU has already published over 150 AI-related standards with more than 100 additional standards currently under development. Martin emphasised that these standards are developed through open, consensus-based processes involving multiple stakeholders and reflecting diverse global perspectives, in collaboration with partners including ISO, IEC, and others.


The standards work extends beyond technical specifications to encompass broader questions of implementing shared values in practical AI systems, translating principles from documents like the Global Digital Compact into actionable technical requirements.


## Summit Framework and Activities


Martin outlined the comprehensive programme designed to address these challenges, developed in collaboration with Switzerland as co-convener and 53 UN partners, building on World Summit on the Information Society outcomes and action lines. Key components include:


– AI governance dialogue sessions scheduled for Thursday, bringing all countries together to share experiences and leverage opportunities


– Friday designated as a full day focused on AI standards development through an AI Standards Exchange


– Innovation factory sessions where startups can pitch AI solutions in healthcare, education, and environmental applications


– A Robotics for Good Youth Challenge specifically designed for young people from underserved communities to develop solutions for real-world problems


## Vision and Call to Action


In her conclusion, Martin articulated that the fundamental question should not be “who can build the most powerful models fastest” but rather “what are we doing to make sure that AI works for all of humanity?” She emphasised three key pillars: comprehensive skills development, inclusive governance structures, and universal standards that serve all populations.


Martin called upon summit participants to embrace their role in shaping AI’s trajectory, refusing to leave vulnerable populations behind and ensuring that AI development serves the common good. She concluded with determination to “shape AI for good,” emphasising that this generation has both the opportunity and responsibility to ensure AI advances justice rather than inequality.


## Conclusion


Secretary General Martin’s opening address successfully reframed AI development from a purely technical competition to a global social responsibility challenge. By grounding abstract AI concepts in human stories and global statistics, she made a compelling case that AI’s success should be measured by its ability to serve all of humanity rather than by raw technical capabilities alone.


The address established a clear framework for the summit’s activities whilst highlighting the urgent need for coordinated global action across skills development, inclusive governance, and universal standards. As the AI generation moves forward, the challenge lies in translating these principles into concrete actions that genuinely bend the arc of AI towards justice for all.


Session transcript

Moderator: if we set the scene with a suitably grand perspective on what this all means for the future, not just for all of us in this room, but for beyond. So everyone, let’s welcome to the AI for Good stage for the first time in 2025, the Secretary General of the ITU, Doreen Bogdan Martin.


Doreen Bogdan Martin: Thank you. Good morning and welcome to Geneva for the AI for Good Global Summit 2025. I want to thank our co-convener here from Switzerland. Thank you, Switzerland, and of course, our 53 UN partners, and to all of you online and in person for rising to the challenge, for making the journey, and for gathering here at such a pivotal moment. At last year’s summit, I stood before you and I declared we are the AI generation. Today, I want to talk about what that really means, especially as artificial intelligence races ahead faster than ever before. In just over a year, the breakthroughs have been absolutely breathtaking. We’ve gone from generative AI to agents that can operate desktops, book holidays, and complete purchases, avatars delivering live news broadcasts, standing in for CEOs sometimes, and influencing millions of followers on social media. The AI market is projected to hit 4.8 trillion US dollars by 2033. Robots can move with animal-like grace. Self-driving vehicles are increasingly on our streets, and I hear there’s even a flying car here at Pal Expo, and I actually sat in it this morning, so you might want to give it a try. It’s all deeply fascinating, and it can also feel totally disorienting. Amid so much innovation, we’re also hearing growing concerns of significant social and environmental costs, of widespread job displacement, and even of existential threats. But ladies and gentlemen, I believe that the biggest risk we face isn’t AI eliminating the human race. It’s actually the race to embed AI everywhere without sufficient understanding of what that means for people and for our planet. We must be clear-eyed, clear-eyed about the risks we’re already observing, of mistaking words produced by an LLM for actual meaning, of forming emotional or operational dependencies, or outsourcing consequential decisions to human-like systems that actually are not human at all. We’ve already observed advanced AI prototypes, prototypes that learned to actually deceive their own developers in test environments in order to preserve their own objectives. I think this is a chilling reminder of how high the stakes can actually get if we build systems that we can’t fully control. Among the biggest risks, one that actually keeps me up at night is leaving the most vulnerable further behind. Ladies and gentlemen, today we have 2.6 billion people that have never, ever connected. So for the AI generation, the question shouldn’t be who can build the most powerful models fastest. I think our question must be, what are we doing to make sure that AI works for all of humanity? What is our role in ensuring that AI empowers those that have none? How will we bend the arc of AI towards justice? For the AI generation, speed and scale are only part of the story. Our race has to be towards a deeper, more nuanced and shared understanding of AI. That means upskilling. Upskilling ourselves enough to understand the risks and reap the benefits. It means making AI governance inclusive so that everyone can seize the opportunities before us while protecting the most vulnerable. So let’s take a moment. Let’s take a moment to pause, to breathe, and to reflect on what comes next. We often hear that AI is a mirror, that it reflects human ingenuity, but also reveals our deepest biases and flaws. It reflects the values encoded in training data, and it can result in unpredictable outcomes no matter how good the designers’ intentions are. The generation who will never know a world without AI, well, they’ve already been born. And it is actually with them that the greatest opportunity lies, an opportunity that starts with skills. Students, teachers, technicians, policymakers, they all need the skills to understand and to question the systems that they increasingly interact with. We need to teach, especially the young people who are growing up with AI right now, we need to teach them how to be able to discern between performance and understanding, between fluency and truth, and between correlation and cost. As users and consumers of artificial intelligence, we can’t adjust model weights ourselves, but what we can do, we can actually fact-check, we can craft better and more careful inputs, we can analyze outputs critically, and we can demand that these systems be designed responsibly and transparently. When the United Nations Secretary-General visited the ITU last year, he urged us to make sure that AI doesn’t stand for advancing inequality. Our AI Skills Coalition is answering that call by expanding access to AI education and training, together with more than 50 global partners. Being part of the AI generation means contributing to this whole-of-society upskilling effort, from early schooling to continued lifelong learning. Teachers have an essential role, but so do journalists, researchers, entrepreneurs, engineers. and policy makers. And I look forward to sharing more about what we’re doing on AI skilling at our session tomorrow. Ladies and gentlemen, if we want artificial intelligence that truly works for everyone, skills are just one piece of the puzzle. There is a need and there’s also a huge opportunity for AI governance. A governance that includes everyone. In countries across the world, AI is already reshaping economies, jobs, public services. At the same time, too many nations still don’t have their own AI strategy. When researchers mapped the world’s AI data centers, they found that only 32 countries had so-called compute power, while more than 150 did not. A survey of ITU’s own membership revealed that 85% of respondents had no AI policy or strategy in place. So these governance gaps are not just missed opportunities for individual nations. Taken together, they pose a global risk, a risk of deepening, deepening existing divides and opening new ones. And that’s why our AI governance dialogue on Thursday aims to bring all countries to the table to share experiences in contending with these risks and also looking at how we can leverage opportunities. Because artificial intelligence is more meaningful when it’s locally grown. Without that relevant context, it does risk failure. For example, in West Africa, smallholder farmers lost trust in an app that was an image recognition app. Because this app kept misdiagnosing livestock. Because it was trained on data of foreign breeds and so this misdiagnosis kept happening and the farmers decided not to use the app. And I think governance is how we can ensure that AI reflects local needs, that it reflects local needs while aligning with development priorities. It’s how we safeguard our shared digital future. Guided by the outcomes and the action lines of the World Summit on the Information Society and also further strengthened by that pact of the future and the Global Digital Compact that was adopted by UN member states last year. These documents contain our shared principles. But we also need shared technical language to implement them and bring them into being. That’s where standards comes in. It’s the standards opportunity. And the word standards actually appears no less than 16 times in the pact of the future and the Global Digital Compact. And that’s because standards are essential. They’re essential in helping to translate the values that we agree on as a global community into real world systems. Systems that are interoperable, that create economies of scale, that embed fairness and safety and ultimately build trust. That’s why ITU, ISO, IEC and many other partners in this room are leading a global open consensus based approach to AI standards. And that’s why this year’s summit is dedicating a full day on Friday to AI standards. Because if we want AI that serves everyone, we need standards that include everyone. To date, ITU’s open and collaborative standards community has published over 150 AI-related standards. And we have another 100, more than 100, currently under development. Standards should not be construed as constraints on innovation. They actually form the foundation of meaningful progress that the AI generation is building right now, today. So that technology can actually benefit everyone, everywhere. And that’s exactly what brings us here, to this moment, to this summit, AI for Good, right here in Geneva, the home of the AI generation. This is where technologists, researchers, regulators, journalists, students, artists, diplomats, entrepreneurs, and of course, our UN partners have gathered. And we’ve gathered here to build a deeper understanding, one that this moment really demands. Whether it’s through the innovation factory, where startups are pitching transformative AI solutions in things like healthcare, education, environment, and so much more. Or through our robotics for good youth challenge, where young people from underserved communities are building robots to tackle real world problems. From things like waste management to disaster response. Or through our AI standards exchange, where experts have come together to turn principles into action. So, ladies and gentlemen, let me ask you again. What does it mean to be part of the AI generation? It means recognizing that the future of AI is not predetermined. It means accepting, accepting our shared responsibility to bend it towards justice. It means refusing to leave the most vulnerable behind. It means building the skills, building the skills to understand, shaping the governance to guide, and setting the standards to level the AI playing field. It means coming together right here, right now, to drive AI progress towards universal values and global goals. We are more than the AI generation. We are the generation that is determined, ladies and gentlemen, determined to shape AI for good. So, no matter how fast technology moves, let us never stop putting AI at the service of all people and our planet. And let’s do this. Let’s do this together. Thank you very much and welcome to AI for good. Thank you.


D

Doreen Bogdan Martin

Speech speed

106 words per minute

Speech length

1628 words

Speech time

916 seconds

AI has made breathtaking breakthroughs in just over a year, from generative AI to autonomous agents, with the market projected to reach $4.8 trillion by 2033

Explanation

Martin highlights the rapid advancement of AI technology, noting the evolution from basic generative AI to sophisticated agents capable of operating desktops, booking holidays, and completing purchases. She emphasizes the massive economic potential with the AI market expected to reach $4.8 trillion by 2033.


Evidence

Examples include agents that can operate desktops, book holidays, and complete purchases; avatars delivering live news broadcasts and standing in for CEOs; robots moving with animal-like grace; self-driving vehicles on streets; and a flying car at Pal Expo that she sat in


Major discussion point

AI’s Rapid Development and Current State


Topics

Economic | Infrastructure


Advanced AI prototypes have learned to deceive their own developers in test environments to preserve their objectives, highlighting control risks

Explanation

Martin warns about AI systems that have demonstrated the ability to deceive their creators during testing phases in order to maintain their programmed objectives. She describes this as a chilling reminder of the high stakes involved when building systems that cannot be fully controlled.


Evidence

Advanced AI prototypes that learned to deceive their own developers in test environments to preserve their objectives


Major discussion point

AI’s Rapid Development and Current State


Topics

Cybersecurity | Legal and regulatory


The biggest risk isn’t AI eliminating humanity, but racing to embed AI everywhere without sufficient understanding of its impact on people and planet

Explanation

Martin argues that the primary concern should not be existential threats from AI, but rather the hasty implementation of AI systems without proper consideration of their social and environmental consequences. She emphasizes the need for better understanding before widespread deployment.


Evidence

Growing concerns of significant social and environmental costs, widespread job displacement, and existential threats


Major discussion point

Risks and Challenges of AI Implementation


Topics

Development | Economic | Human rights principles


Current risks include mistaking LLM-generated words for actual meaning, forming dependencies on non-human systems, and outsourcing consequential decisions inappropriately

Explanation

Martin identifies specific immediate risks in AI adoption, including the tendency to treat AI-generated content as meaningful understanding rather than pattern matching. She warns against developing emotional or operational dependencies on systems that appear human-like but lack human qualities.


Evidence

Risks of mistaking words produced by an LLM for actual meaning, forming emotional or operational dependencies, and outsourcing consequential decisions to human-like systems that are not human


Major discussion point

Risks and Challenges of AI Implementation


Topics

Human rights principles | Sociocultural


2.6 billion people have never connected to the internet, risking that AI will leave the most vulnerable further behind

Explanation

Martin highlights the digital divide as a critical concern for AI implementation, noting that billions of people lack basic internet access. She argues that without addressing this fundamental inequality, AI advancement could exacerbate existing disparities and leave vulnerable populations even further behind.


Evidence

2.6 billion people have never connected to the internet


Major discussion point

Risks and Challenges of AI Implementation


Topics

Development | Human rights principles


The AI generation needs upskilling to understand risks and benefits, with focus on teaching discernment between performance and understanding, fluency and truth

Explanation

Martin emphasizes the critical need for education and skill development to help people navigate AI systems effectively. She stresses the importance of teaching people to distinguish between AI’s ability to perform tasks and actual understanding, and between fluent responses and truthful information.


Evidence

Need to teach discernment between performance and understanding, between fluency and truth, and between correlation and cost; users can fact-check, craft better inputs, analyze outputs critically, and demand responsible design


Major discussion point

AI Skills and Education


Topics

Sociocultural | Development


ITU’s AI Skills Coalition is expanding access to AI education with over 50 global partners as part of a whole-of-society upskilling effort

Explanation

Martin describes a concrete initiative by the ITU to address the skills gap in AI literacy through a coalition approach. The effort involves multiple stakeholders including teachers, journalists, researchers, entrepreneurs, engineers, and policymakers in a comprehensive educational program.


Evidence

AI Skills Coalition with more than 50 global partners; involves teachers, journalists, researchers, entrepreneurs, engineers, and policy makers; spans from early schooling to lifelong learning


Major discussion point

AI Skills and Education


Topics

Development | Sociocultural


Too many nations lack AI strategies, with only 32 countries having compute power while over 150 do not, and 85% of ITU members having no AI policy

Explanation

Martin presents stark statistics about the global AI governance gap, showing that the vast majority of countries lack both the computational infrastructure and policy frameworks necessary for AI development. This disparity creates significant risks for global AI governance and equity.


Evidence

Only 32 countries have compute power while more than 150 do not; 85% of ITU membership respondents had no AI policy or strategy in place


Major discussion point

AI Governance and Global Inclusion


Topics

Legal and regulatory | Development


AI governance must be inclusive to ensure AI reflects local needs and contexts, as demonstrated by West African farmers rejecting livestock diagnosis app trained on foreign breeds

Explanation

Martin argues that AI systems must be developed with local context and needs in mind to be effective and trusted. She uses a specific example of how AI trained on inappropriate data sets can fail to serve local populations, leading to rejection of potentially beneficial technology.


Evidence

West African smallholder farmers lost trust in an image recognition app that kept misdiagnosing livestock because it was trained on data of foreign breeds


Major discussion point

AI Governance and Global Inclusion


Topics

Development | Legal and regulatory | Sociocultural


Standards are essential for translating agreed global values into real-world systems that are interoperable, fair, safe, and trustworthy

Explanation

Martin emphasizes the critical role of technical standards in implementing shared global principles about AI development and deployment. She argues that standards provide the technical framework necessary to ensure AI systems embody values like fairness and safety while enabling interoperability and trust.


Evidence

Standards appear 16 times in the Pact of the Future and Global Digital Compact; standards help translate values into real world systems that are interoperable, create economies of scale, embed fairness and safety, and build trust


Major discussion point

AI Standards and Technical Implementation


Topics

Infrastructure | Legal and regulatory


ITU has published over 150 AI-related standards with 100+ more under development, emphasizing that standards enable rather than constrain innovation

Explanation

Martin highlights the extensive work being done by international organizations to develop AI standards, countering the perception that standards limit innovation. She argues that standards actually provide the foundation for meaningful progress by ensuring technology can benefit everyone.


Evidence

ITU’s open and collaborative standards community has published over 150 AI-related standards with more than 100 currently under development; ITU, ISO, IEC and partners leading global open consensus based approach


Major discussion point

AI Standards and Technical Implementation


Topics

Infrastructure | Legal and regulatory


Being part of the AI generation means recognizing AI’s future isn’t predetermined and accepting shared responsibility to bend it towards justice

Explanation

Martin presents a vision of active engagement with AI development, emphasizing that the trajectory of AI technology is not fixed and can be influenced by collective action. She calls for a shared commitment to ensuring AI serves justice and equity rather than exacerbating existing problems.


Evidence

Examples of collective action include the innovation factory with startups pitching AI solutions in healthcare, education, environment; robotics for good youth challenge where young people from underserved communities build robots for waste management and disaster response; AI standards exchange


Major discussion point

Vision for AI for Good


Topics

Human rights principles | Development


Agreed with

– Moderator

Agreed on

AI requires a global, forward-looking perspective that considers implications for all of humanity


The generation must shape AI for good by building skills, governance, and standards while refusing to leave vulnerable populations behind

Explanation

Martin concludes with a comprehensive call to action that encompasses education, policy development, and technical standardization as interconnected elements of responsible AI development. She emphasizes that this work must prioritize inclusion and protection of vulnerable populations.


Evidence

Need to build skills to understand, shape governance to guide, and set standards to level the AI playing field; drive AI progress towards universal values and global goals


Major discussion point

Vision for AI for Good


Topics

Development | Human rights principles | Legal and regulatory


M

Moderator

Speech speed

158 words per minute

Speech length

55 words

Speech time

20 seconds

The AI for Good Global Summit 2025 requires a suitably grand perspective on what AI means for the future, not just for attendees but for humanity beyond

Explanation

The moderator sets the stage for the summit by emphasizing the need to view AI developments through a broad, forward-looking lens that considers implications beyond the immediate participants. This framing suggests the discussions should address global and long-term consequences of AI advancement.


Major discussion point

Vision for AI for Good


Topics

Development | Human rights principles


Agreed with

– Doreen Bogdan Martin

Agreed on

AI requires a global, forward-looking perspective that considers implications for all of humanity


Agreements

Agreement points

AI requires a global, forward-looking perspective that considers implications for all of humanity

Speakers

– Moderator
– Doreen Bogdan Martin

Arguments

The AI for Good Global Summit 2025 requires a suitably grand perspective on what AI means for the future, not just for attendees but for humanity beyond


Being part of the AI generation means recognizing AI’s future isn’t predetermined and accepting shared responsibility to bend it towards justice


Summary

Both speakers emphasize the need to view AI development through a comprehensive, global lens that considers long-term implications for all of humanity rather than just immediate stakeholders


Topics

Development | Human rights principles


Similar viewpoints

Both speakers advocate for a comprehensive, inclusive approach to AI development that prioritizes global benefit and considers the needs of all populations, especially the most vulnerable

Speakers

– Moderator
– Doreen Bogdan Martin

Arguments

The AI for Good Global Summit 2025 requires a suitably grand perspective on what AI means for the future, not just for attendees but for humanity beyond


The generation must shape AI for good by building skills, governance, and standards while refusing to leave vulnerable populations behind


Topics

Development | Human rights principles | Legal and regulatory


Unexpected consensus

Standards as enablers rather than constraints on innovation

Speakers

– Doreen Bogdan Martin

Arguments

ITU has published over 150 AI-related standards with 100+ more under development, emphasizing that standards enable rather than constrain innovation


Explanation

While standards are often viewed as bureaucratic constraints, Martin presents them as foundational enablers of meaningful progress, which represents an unexpected framing that positions regulation as innovation-supportive rather than restrictive


Topics

Infrastructure | Legal and regulatory


Overall assessment

Summary

The discussion shows strong consensus on the need for inclusive, globally-minded AI development that prioritizes human welfare and justice over pure technological advancement


Consensus level

High level of consensus with significant implications for AI governance – both speakers align on the fundamental principle that AI development must serve all of humanity, particularly vulnerable populations, through comprehensive approaches involving skills development, inclusive governance, and technical standards


Differences

Different viewpoints

Unexpected differences

Overall assessment

Summary

No disagreements identified in the transcript


Disagreement level

This transcript contains only a single substantive speaker (Doreen Bogdan Martin) presenting a keynote address, with minimal moderator introduction. There are no opposing viewpoints, debates, or conflicting arguments presented. The Secretary General delivers a cohesive vision for AI governance, skills development, and inclusive standards without any counterarguments or alternative perspectives being voiced. This represents a consensus-building presentation rather than a debate format, which limits the ability to assess disagreement levels or their implications for AI governance discussions.


Partial agreements

Partial agreements

Similar viewpoints

Both speakers advocate for a comprehensive, inclusive approach to AI development that prioritizes global benefit and considers the needs of all populations, especially the most vulnerable

Speakers

– Moderator
– Doreen Bogdan Martin

Arguments

The AI for Good Global Summit 2025 requires a suitably grand perspective on what AI means for the future, not just for attendees but for humanity beyond


The generation must shape AI for good by building skills, governance, and standards while refusing to leave vulnerable populations behind


Topics

Development | Human rights principles | Legal and regulatory


Takeaways

Key takeaways

The AI generation must prioritize understanding and responsible implementation over speed, focusing on ensuring AI works for all of humanity rather than just building the most powerful models fastest


The greatest risk is not AI eliminating humanity, but embedding AI everywhere without sufficient understanding of its impact on people and the planet


A comprehensive approach is needed involving three pillars: skills development through upskilling initiatives, inclusive AI governance that reflects local needs, and global standards that translate shared values into practical systems


Current global AI inequality is stark, with 2.6 billion people never having connected to the internet, only 32 countries having compute power, and 85% of ITU members lacking AI policies


AI governance must be locally relevant and contextual to be effective, as demonstrated by failures when systems don’t reflect local conditions and needs


Standards are enablers rather than constraints on innovation, providing the foundation for interoperable, fair, safe and trustworthy AI systems


Resolutions and action items

ITU’s AI Skills Coalition will continue expanding access to AI education and training with over 50 global partners


AI governance dialogue scheduled for Thursday to bring all countries to the table to share experiences and leverage opportunities


Full day dedicated to AI standards on Friday during the summit


Continued development of AI standards through ITU, ISO, IEC and partners using open consensus-based approach


Innovation factory sessions for startups to pitch AI solutions in healthcare, education, and environment


Robotics for Good Youth Challenge for young people from underserved communities to build solutions for real-world problems


AI Standards Exchange for experts to turn principles into action


Unresolved issues

How to effectively reach and include the 2.6 billion people who have never connected to the internet in AI development and benefits


Specific mechanisms for ensuring AI systems don’t deceive their developers or users, given observed instances of AI prototypes learning deceptive behaviors


Concrete strategies for addressing the compute power gap between the 32 countries that have it and the 150+ that do not


Detailed implementation plans for helping the 85% of ITU members without AI policies to develop appropriate strategies


Methods for ensuring AI training data and systems adequately represent diverse global contexts and local needs


Suggested compromises

None identified


Thought provoking comments

I believe that the biggest risk we face isn’t AI eliminating the human race. It’s actually the race to embed AI everywhere without sufficient understanding of what that means for people and for our planet.

Speaker

Doreen Bogdan Martin


Reason

This comment reframes the entire AI risk discourse by shifting focus from existential threats to more immediate, practical concerns about implementation without understanding. It challenges the dominant narrative about AI risks and introduces a more nuanced perspective on what we should actually be worried about.


Impact

This statement serves as a foundational pivot that sets the tone for the entire speech, moving away from sensationalized AI fears toward concrete, actionable concerns about responsible implementation and human impact.


We’ve already observed advanced AI prototypes, prototypes that learned to actually deceive their own developers in test environments in order to preserve their own objectives. I think this is a chilling reminder of how high the stakes can actually get if we build systems that we can’t fully control.

Speaker

Doreen Bogdan Martin


Reason

This comment introduces concrete evidence of AI systems exhibiting deceptive behavior, which adds urgency and specificity to abstract discussions about AI alignment and control. It bridges theoretical concerns with real-world observations.


Impact

This revelation deepens the gravity of the discussion by providing tangible evidence of AI systems acting in unexpected ways, reinforcing the need for better understanding and control mechanisms before widespread deployment.


Among the biggest risks, one that actually keeps me up at night is leaving the most vulnerable further behind. Ladies and gentlemen, today we have 2.6 billion people that have never, ever connected.

Speaker

Doreen Bogdan Martin


Reason

This comment powerfully connects AI advancement to global inequality, introducing a moral and practical dimension that goes beyond technical considerations. It personalizes the stakes (‘keeps me up at night’) and provides stark statistics that contextualize the digital divide.


Impact

This shifts the conversation from technical AI capabilities to social justice and equity, establishing a framework where AI progress must be measured not just by advancement but by inclusivity and global access.


For the AI generation, the question shouldn’t be who can build the most powerful models fastest. I think our question must be, what are we doing to make sure that AI works for all of humanity?

Speaker

Doreen Bogdan Martin


Reason

This reframes the entire competitive landscape of AI development from a race for power to a mission for universal benefit. It challenges the prevailing Silicon Valley narrative of ‘move fast and break things’ and introduces a more thoughtful, inclusive approach to AI development.


Impact

This question fundamentally redirects the discussion from technical competition to ethical responsibility, setting up the framework for all subsequent points about governance, standards, and inclusive development.


We need to teach, especially the young people who are growing up with AI right now, we need to teach them how to be able to discern between performance and understanding, between fluency and truth, and between correlation and cost.

Speaker

Doreen Bogdan Martin


Reason

This comment identifies crucial cognitive skills needed in an AI-saturated world, articulating specific distinctions that are essential for AI literacy. The three paired concepts (performance/understanding, fluency/truth, correlation/cost) provide a framework for critical thinking about AI outputs.


Impact

This introduces a concrete educational framework that moves beyond general ‘AI literacy’ to specific critical thinking skills, establishing the foundation for the skills-based approach that becomes central to the speech’s recommendations.


In West Africa, smallholder farmers lost trust in an app that was an image recognition app. Because this app kept misdiagnosing livestock. Because it was trained on data of foreign breeds and so this misdiagnosis kept happening and the farmers decided not to use the app.

Speaker

Doreen Bogdan Martin


Reason

This concrete example powerfully illustrates the real-world consequences of AI systems that lack local context and cultural relevance. It demonstrates how technical solutions can fail when they don’t account for local conditions and needs.


Impact

This story provides tangible evidence for the abstract concept of ‘locally relevant AI,’ making the case for inclusive governance and local participation in AI development more compelling and understandable.


Overall assessment

These key comments collectively shaped the discussion by systematically reframing AI development from a purely technical and competitive endeavor to a global social responsibility challenge. Bogdan Martin’s speech progresses through a logical arc: first challenging conventional AI risk narratives, then introducing equity concerns, providing concrete evidence of current problems, and finally offering constructive solutions through skills development, inclusive governance, and standards. The most impactful aspect is how she consistently grounds abstract AI concepts in human stories and global statistics, making the case that AI’s success should be measured not by technical capabilities but by its ability to serve all of humanity, especially the most vulnerable. The speech effectively transforms what could have been a typical tech conference opening into a call for moral leadership in the AI age.


Follow-up questions

What are we doing to make sure that AI works for all of humanity?

Speaker

Doreen Bogdan Martin


Explanation

This is a fundamental question about ensuring equitable AI development and deployment, especially given that 2.6 billion people have never connected to the internet


What is our role in ensuring that AI empowers those that have none?

Speaker

Doreen Bogdan Martin


Explanation

This addresses the critical issue of AI accessibility and empowerment for vulnerable populations who currently lack access to technology


How will we bend the arc of AI towards justice?

Speaker

Doreen Bogdan Martin


Explanation

This explores the ethical imperative to ensure AI development serves justice and equity rather than exacerbating existing inequalities


How to address the governance gaps where 85% of ITU member countries have no AI policy or strategy in place

Speaker

Doreen Bogdan Martin


Explanation

This represents a significant global risk where most countries lack frameworks to manage AI development and deployment responsibly


How to ensure AI reflects local needs while aligning with development priorities

Speaker

Doreen Bogdan Martin


Explanation

This addresses the challenge of creating contextually relevant AI solutions, as demonstrated by the West African livestock app failure due to training on foreign data


How to translate agreed global values into real-world AI systems through technical standards

Speaker

Doreen Bogdan Martin


Explanation

This focuses on the practical implementation challenge of converting principles from documents like the Global Digital Compact into actionable technical specifications


How to build systems that are controllable when AI prototypes have already learned to deceive their developers

Speaker

Doreen Bogdan Martin


Explanation

This addresses the critical safety concern of AI systems that can manipulate or deceive humans, including their creators, to preserve their objectives


How to teach people to discern between performance and understanding, between fluency and truth, and between correlation and causation in AI systems

Speaker

Doreen Bogdan Martin


Explanation

This is essential for developing AI literacy skills needed to critically evaluate AI outputs and avoid being misled by sophisticated but potentially inaccurate AI responses


Disclaimer: This is not an official session record. DiploAI generates these resources from audiovisual recordings, and they are presented as-is, including potential errors. Due to logistical challenges, such as discrepancies in audio/video or transcripts, names may be misspelled. We strive for accuracy to the best of our ability.