Doreen Bogdan-Martin
Secretary-General, International Telecommunication Union (ITU)
Keynote at the
AI for Good Global Summit 2026
Together for AI for Good
[As prepared for delivery]
Good morning, and welcome to the AI for Good Global Summit!
Yesterday, right here at Palexpo, leaders from more than 160 countries concluded two-day discussions on how to govern AI for the benefit of all humanity at the first-ever UN Global Dialogue on AI Governance, which ITU was proud to co-coordinate with UNESCO (United Nations Educational, Scientific and Cultural Organization).
Also, for the first time, the Independent International Scientific Panel on AI laid its initial evidence on the table.
These contributions provide an important waypoint for the road ahead.
Through the Global Dialogue, this Summit, and the WSIS Forum, the UN (United Nations) system is convening stakeholders from across sectors to exchange experiences, and deepen cooperation, as we move forward together from conversation, to implementation, to impact.
Because no single stakeholder can shape the future of AI alone.
It needs builders - it needs
you.
And if anyone can help us do this, it's the amazing AI for Good community, right here in this room.
AI for Good was born right here in Geneva, in 2017, based on our conviction that artificial intelligence (AI), deployed responsibly, could help solve humanity's most pressing problems - from hunger, to disease, to a warming planet.
Today, that idea is being tested, including by the challenges AI itself is bringing, even as we strive to use it for good.
The resources AI is burning through with electricity use on track to more than double by 2030.
The workforces being disrupted; the growing divides.
With developed countries adopting AI more than twice as fast as in the developing world, and with 2.2 billion people still completely offline.
At the same time, AI is testing the resilience of our economies, of our security networks - even our health and well-being.
This shift has to do with a word we have heard over and over: Agent.
AI agents that conduct "deep research" or build applications while you sleep.
Whose traffic has overtaken human activity on the Internet.
AI agents represent an exciting step forward - extending human capabilities - and expanding possibilities for innovation in science, education, healthcare, and so much more.
At the same time, agents may soon be able to not only edit their own code, but even build their own AI agents, with little or no human oversight.
Some warn that the next AI system humanity builds could be the last one we are ever able to control.
The word agency comes from the Latin
agere. To set something in motion. To take action. To shape outcomes.
And yes, AI systems are increasingly able to act on our behalf with growing autonomy.
But agency in the fullest sense is also the ability to exercise judgment, discernment and moral responsibility.
And guess what?
That's where all of you - all of us - come in.
Our agency is the power to keep humanity in the loop.
To decide whether, when, and how AI acts on our behalf.
To stay in control.
Not to fear AI. But to shape it.
That's what agency is about.
Because technology only has agency if and when we humans decide to give it.
So here is what I want you to do - especially after you leave this Summit.
Use that power. Use your AI agency. And use it for good.
Let me show you what that looks like.
Starting with solutions.
Imagine that an English-speaking mother-to-be asks a popular LLM (large language model) about swollen legs late in her pregnancy.
She'd likely be warned about pre-eclampsia, a complication that kills more than 70,000 mothers a year.
Now imagine another expectant mother asks the very same question in Kiswahili.
Because AI systems often underperform in underrepresented languages, she might be given a different answer or even be told not to worry.
In Lagos, Nigeria, a young researcher and engineer named Israel Adegoke used his agency to tackle this problem.
He built his own speech recognition system that can run offline, on low-resource devices, and carefully fine-tuned it in Kiswahili, one of the most widely-spoken African languages.
His natural language model for health questionnaires won the AI and Machine Learning (ML) challenge led by ITU and Zindi.
But to use agency to build human-centric AI like Israel Adegoke did, you also need skills.
That's why we launched the AI Skills Coalition, A global platform with 70 partners offering more than 200 courses in 13 languages.
And that's why, together with will.i.am and Google, ITU is training thousands of students and teachers in AI and robotics, starting with five countries.
Some of those young people will take their skills to the Robotics for Good Youth Challenge, where, since 2024, students in more than 50 countries have built their own robots to tackle real-world problems.
Last year, teams from Venezuela and Brazil won the Disaster Response Challenge. And later this week, we will meet the winners of the Food Security Challenge with FAO (Food and Agriculture Organization).
That's human agency too - deciding what robots should and should not do - and especially teaching our children these skills, as AI becomes increasingly present in their lives.
As a mother, I feel this responsibility personally and deeply.
It is a matter of absolute urgency, which is why ITU is partnering with stakeholders worldwide, to ensure kids are protected, empowered and safe in an increasingly AI-driven world.
We must do everything we can to support young people, and all people, to understand AI's risks and limits, and when to refuse outsourcing decisions that must stay human.
As autonomous systems spread into hospitals, classrooms, courts, banks and government services - human agency is how we stay in control.
Agency is also something we build together, through standards.
A standard is how you take a value and build it into the way AI actually operates. How you turn it into transparency, trust, and accountability.
Think about standards born at ITU - like telephone numbers and digital certificates for trusted exchanges online.
And now think of the AI agents in our hospitals, banks and other critical infrastructure - buying, selling, and acting on our behalf.
They don't have phone numbers, passports or ID numbers.
Who are they?
That is one of the most fundamental questions in AI security today.
And the consequences of getting it wrong are more than a dropped call.
That's why ITU is convening standards experts from industry, governments, the technical community, and academia - to develop the standards needed for AI agents to be trustworthy and verifiable - the identity layer of the age of artificial intelligence.
Giving AI agents trusted identities will help unlock their massive potential, allowing them to collaborate with us humans safely across critical domains, giving us confidence in what we are interacting with.
Returning to that word: Agency. The power to stay in control of and take responsibility for what AI does.
Earlier this year, Pope Leo XIV wrote that “no one can single-handedly bear the weight of the challenges the world is facing."
We all have our own areas for action to orient this powerful technology toward the common good.
“Let each builder choose with care how to build," he tells us in
Magnifica Humanitas.
Whether or not we have access to a frontier model, we all have AI agency.
We can all start somewhere, whether it's upskilling ourselves or others, working on an AI security standard, or enacting a policy solution that safeguards people and our planet.
If we get this right, AI will not replace human agency. It will amplify it.
The new AI for Good Global Commission, and the AI for Good community of more than 60,000 members in 180 countries and counting - give us 60,000 reasons to be hopeful.
You are why I am hopeful that we will, all of us, use our collective AI agency right here, right now, and always for good.
