Artificial intelligence is developing faster than many governments, institutions and communities can adapt to it. The question is no longer simply whether people will use AI, but who will have the power, knowledge and resources to shape how it develops. This question was at the heart of the first United Nations Global Dialogue on Artificial Intelligence Governance, held in Geneva on 6–7 July 2026.
The Dialogue brought together more than 3,000 participants from 163 countries, including governments, technology companies, civil society, researchers, international organisations and other stakeholders. Its purpose was to create a space for countries and stakeholders to discuss how international cooperation on AI governance can develop.
One of the strongest messages emerging from the Dialogue was that AI presents both enormous opportunities and serious challenges. Participants discussed its potential to contribute to healthcare, education, agriculture, scientific research, public services and sustainable development. At the same time, they highlighted concerns around human rights, employment, cybersecurity, misinformation, environmental sustainability, child safety and the concentration of technological power.
The AI divide is about more than internet access
Perhaps one of the most important discussions was about the growing AI divide.
Countries do not enter the AI era with the same infrastructure, computing power, data, technical expertise or institutional capacity. Some countries are developing advanced AI systems, while others are primarily consumers of technologies developed elsewhere.
The Dialogue therefore emphasized that simply giving people access to AI tools is not enough. Countries need the capacity to develop, adapt, govern and benefit from AI themselves. This includes investment in connectivity, computing infrastructure, skills, research, data governance and institutions.
This distinction is particularly important for developing countries. Without investment in local capacity, there is a risk that countries will remain dependent on technologies, models and infrastructure developed elsewhere, while having limited influence over the rules governing them.
The conversation therefore needs to move from “Who has access to AI?” to a broader question: “Who has the capacity to shape AI?”
Human rights must remain at the centre
Another major theme was the relationship between AI governance and human rights.
Participants emphasized that international human rights law provides an important foundation for governing AI. Transparency, accountability, non-discrimination, privacy and meaningful human oversight were repeatedly highlighted as important throughout the AI life cycle.
This becomes even more important as AI systems become increasingly autonomous. The Dialogue identified the need for clear responsibility when AI systems cause harm, mechanisms for people to challenge decisions, independent testing and auditing, and human decision-makers who can intervene when necessary.
In other words, technological capability should not remove human responsibility.
Children and women cannot be an afterthought
The Dialogue also placed considerable attention on groups that can experience particular risks from AI.
Children were identified as requiring specific protection as their use of AI tools grows. Participants called for child-rights approaches to be incorporated into the design, development and deployment of AI systems.
Gender was another important part of the conversation. Participants highlighted the disproportionate effects that some AI-enabled harms can have on women and girls, alongside their underrepresentation in AI development and governance. They called for greater participation of women in policymaking and standard-setting, as well as better gender-disaggregated data.
This matters because AI systems are shaped by the data, people and environments in which they are developed. If particular communities are absent from these processes, their experiences and needs can also be overlooked.
AI needs to understand more than the world's dominant languages
Language and culture emerged as another significant governance issue.
Many AI systems are trained primarily using data from dominant languages and cultural contexts. The Dialogue recognized that this can leave many communities poorly represented in the digital AI ecosystem.
Participants therefore emphasized linguistic and cultural inclusion by design, including support for smaller, locally adapted models and datasets representing low-resource and Indigenous languages.
For countries and communities whose languages are poorly represented online, this is not simply a technical issue. It affects whether people can meaningfully participate in an increasingly AI-mediated digital world.
The world may not need one AI law
Interestingly, the Dialogue did not suggest that every country needs exactly the same AI regulatory framework.
Instead, participants discussed interoperability: finding ways for different national and regional governance systems to work together through shared standards, definitions, risk classifications, testing procedures and benchmarks.
This reflects the reality that countries have different legal systems, economic circumstances and policy priorities. The challenge is therefore not necessarily to create one identical global regulatory system, but to find areas where cooperation and compatibility are possible.
AI governance is also climate governance
The environmental consequences of AI received attention as well.
AI infrastructure requires significant amounts of energy, water and materials, and participants argued that these impacts should be measured and transparently disclosed. Environmental sustainability was therefore framed as a core AI governance issue rather than something separate from discussions about technology.
This raises an important question for the future: how can societies benefit from AI while also understanding and managing the environmental resources required to build and operate increasingly powerful systems?
From principles to action
Perhaps the biggest challenge now is moving from discussion to implementation.
The Co-Chairs' Summary makes clear that the themes emerging from the Dialogue are not negotiated outcomes or consensus positions. Instead, they represent recurring issues raised by participants and areas for continued discussion.
That distinction matters. The Dialogue has not created a single global AI rulebook. What it has done is create a platform where governments and other stakeholders can identify shared concerns, compare approaches and explore opportunities for cooperation.
The AI Dialogue Partnerships Hub also provides a practical avenue for this work, showcasing more than 350 commitments, announcements and initiatives related to areas such as AI literacy, capacity building, safety, transparency, infrastructure, health, workforce development and inclusion.
What comes next?
The first UN Global Dialogue has established a starting point, but the harder work is ahead.
If AI is going to benefit people everywhere, countries need more than access to AI products. They need the infrastructure, skills, data, institutions and decision-making power necessary to participate in shaping the technology.
The Dialogue also points to the importance of ensuring that women, young people, developing countries, civil society and affected communities are not simply consulted after major decisions have already been made. They need meaningful opportunities to participate in the processes that determine how AI is developed and governed.
The next Global Dialogue on AI Governance will take place in New York in May 2027. The period between the two meetings will be important for turning the ideas discussed in Geneva into practical cooperation, stronger capacity and concrete governance mechanisms.
The future of AI governance will not be determined only by how powerful AI becomes. It will also depend on whether the institutions governing it can keep pace, whether countries have the capacity to participate, and whether the people most affected by AI have a meaningful voice in shaping its future.
Comments
Post a Comment