In September, UN Global Pulse and the UN 2.0 Digital and Data Community launched a new event series, Beyond the Talking Points, exploring how priorities emerging from global debates on digital technologies are interpreted, operationalized, and applied across organizations, products, governance processes, and partnerships. The focus is not on which principles matter, but on how they take shape in practice, where implementation challenges emerge, and where meaningful cooperation is already happening.
The series kicked-off with a reflection on the inaugural AI Governance Dialogue, which brought together governments, industry, academia, civil society, and international organizations to discuss international AI governance.
The strongest overall message from this first session was that the first Global Dialogue was an important first step. Still, the real test is what happens when global principles meet real-world systems, institutions and communities. This came through clearly when one participant connected several seemingly different contributions — from safety and accountability to participation and country-level implementation — to the same underlying question: “Does this framework hold once it meets the ground it’s meant to govern?”
The conversation moved beyond whether AI should be governed toward how governance can become practical, measurable, inclusive and responsive enough to keep pace with rapidly evolving AI systems.
Some key messages from the conversation and audience participation
1. AI governance cannot wait for perfect evidence
Governance needs to become more anticipatory and adaptive. Waiting until evidence is complete or technologies have stabilised risks letting today’s deployments set tomorrow’s norms by default. Participants highlighted that AI is advancing rapidly, with one participant stressing that “AI is advancing faster than governance”, while another cautioned against waiting for extensive evidence before taking policy decisions. The latter argued that policymakers need to act both to harness opportunities and mitigate risks, even when economic and other evidence may take years to emerge. This was sharpened by discussion of agentic AI – an artificial intelligence system that can accomplish a specific goal with limited supervision or human oversight – with one participant arguing that agentic systems are already being deployed at scale while governance mechanisms remain unresolved, creating a widening gap between the “capability curve” and the “governance curve.”
2. Trust needs to become something we can actually measure
One of the strongest substantive threads concerned trustworthiness. A particularly useful question was raised: while organisations can measure model performance, what should they measure to determine whether an AI system deserves to be trusted in practice? This matters especially when AI informs decisions such as displacement forecasting or humanitarian response. Participants strongly feel that trust cannot be reduced to whether an AI system performs well on average. It requires lifecycle verification, context-specific testing, monitoring for drift, transparency about failures and meaningful human oversight.
3. Interoperability is not just a technical issue—it is a governance issue
Interoperability emerged as both a practical near-term governance objective and a potential foundation for addressing deeper questions of market concentration, portability, user choice and accountability. Participants steered the discussion toward data portability and the ability to switch between AI systems, arguing that users should not become locked into a particular provider’s AI ecosystem. This was then connected to the broader question of whether interoperability is merely the most realistic short-term objective or whether it could eventually lead toward binding global rules, potentially involving the UN, regional institutions such as the EU and the private sector.
4. AI governance must include the people who build and use the systems
Effective AI governance needs meaningful participation from technical practitioners, developers, users, civil society and communities — not only policymakers and institutions. The point resonated strongly during the conversation. The underlying message was that governance frameworks cannot remain lengthy, abstract documents developed separately from the technical people who will have to implement them.
This connects closely to another major concern: who gets to participate in AI governance? The discussion questioned the composition of the Global Dialogue and the absence of many private-sector actors. One participant highlighted the concentration of AI development in the US and China and noted the predominance of private-sector models, while also pointing to the limited participation of these actors in the Dialogue.
5. The global conversation has to connect to very different local realities
This was perhaps the most relevant message for the UN system. Participants highlighted the challenge of translating global governance discussions into practice across 130+ very different country contexts, pointing specifically to information integrity, trust, accountability, interoperability, inclusion and the AI divide. The importance of multilingual and local-context dimensions of responsible AI was also stressed.
This was reinforced by discussion about AI systems working in languages and contexts that vendors may not prioritise. One participant noted that existing feedback mechanisms often do not provide public evidence about whether AI tools work for people, particularly in underrepresented languages and contexts. At the end of the day, it was clear that global AI governance will only be meaningful if it can travel from the global level to the local level—and still work.
6. Inclusion needs to go beyond access
An inclusive AI future is not only about who can access AI. It is also about who benefits, whose jobs and livelihoods are affected, whose languages and experiences are represented, and who bears the environmental and social costs. The conversation broadened the meaning of inclusion considerably.
Participants pointed to multiple dimensions of accessibility, including physical disability, neurodivergence, financial inclusion and technical literacy, while asking how AI can empower vulnerable and lower-income communities. They also raised concerns about the environmental footprint of AI infrastructure, employment displacement and the macroeconomic consequences of automation. The proposed alternative was a stronger focus on human–AI collaboration and augmentation rather than unchecked automation.
7. Information integrity, privacy and human agency remain central
There were also several important practical concerns from participants working across the UN system. Participants raised the challenge of using generative AI for images, design and communications, particularly around quality control, brand integrity, credibility and ethical risks.
Concerns around AI-generated emails potentially eroding people’s personality and authenticity, alongside questions about privacy and uncontrolled uploading of documents, were also raised. These contributions point toward an important operational dimension of governance: organisations need practical guidance for responsible AI use, not just high-level principles.
Conclusion – why does this all matter?
AI governance is critical now more than ever. As artificial intelligence and other emerging technologies advance rapidly, we need stronger regulatory frameworks to curtail risks, keep technology companies accountable, prevent the AI divide from widening further, and ensure the technology serves people and planet for good. And as AI systems increasingly absorb and reproduce human knowledge, governance must also consider whose knowledge is represented, whose knowledge is preserved, and whose local and community perspectives risk being lost.
If you missed this webinar, you can access the recording and resources.
Don’t miss our next webinar on Advancing Open and Interoperable Data Across the UN. This session will spotlight the UN System Data Commons as a practical example of digital cooperation in action. The platform serves as a single point of entry for finding, accessing, exploring, and analyzing official UN system statistics relevant to global development priorities, including the Sustainable Development Goals (SDGs).
Register for the session here!