2026 Global Dialogue on AI Governance and other events: More Insights

AI literacy is about understanding AI – what AI is and is not. What AI can and cannot do and what it should and should not do.

1. Perception of AI History: There is a perception that the AI age began three years ago with ChatGPT. This is not true, but during the pandemic, there was rapid onboarding of the technology by governments, the sector, and individuals. 

2. Metaphors not accurate: AI is like electricity or like a plane (It is NOT).

Electricity does not steal your copyrighted work or your livelihood – AI can. A plane cannot move across sectors – say from aviation to medicine – AI can.

3. AI systems definitions are not clear: Even those adopted by OECD and the EU do not look to the future when we may not use silicon-based AI systems but other materials like carbon-based AI systems (there are experiments with brain bots and DNA storage going on). Of course, governance is lagging.

4. Confusion on what is meant by “intelligence”:  We know less about the brain than you think and hence about human intelligence. What we believe about how the brain works is a field that is still evolving! The original definition of AI focused on getting machines to “mimic” human intelligence. Did you know your heart has neurons (things we thought were only in the brain?) and your stomach influences your brain through 100 million nerve cells? So, no, an AI neural network is not equivalent to our human brain and will not be for some time, as our brain and body do so much with so little energy consumption (20 watts for the Brain and 80 Watts for the Body). Think about what your brain and body can do versus a laptop that is NOT connected to the internet (a chatbot/GenAI query could be 4 Watts per request). 

5. Confusion on governing AI: Some people thought governance was the government’s responsibility. It is actually complex – international, regional, and national government frameworks, strategies, and regulations (including data, security, trade, industry, and liability laws), Standards (such as ISO, IEC, and IEEE), and the responsibilities of the individual (as set out in the Universal Declaration of Human Rights, Article 29). There was a lot of debate on the power some countries and organizations had in terms of AI frontier models and the impact it would have on other countries and individuals.

Some people thought governance was all about models (algorithms and weights), language inclusiveness, and data. To create AI systems, you need Digital Public Infrastructure (DPI), hardware, software, data, and human beings to drive the system's motives and values. I came from a recent conference on international business and heard several professors speak of data as complementary. Data is essential to AI, which is why there are so many lawsuits over copyright infringement and new protections for children on social media. If you underestimate the cost of AI (hardware that needs to be upgraded every 5 years), software that needs to be updated (and all the license costs), and the increasing cost of cybersecurity (it will go up every year, the more we connect systems together and the more data it has) – you underestimate your ability to govern AI.

Way forward

1. AI literacy – which is different from how to use AI to understanding AI, its responsibility and ethics.

2. AI governance needs operationalization, implementation, and measurement. There was some discussion about the baseline – should we hold governments accountable to the frameworks they have endorsed? What would be great measurement standards? Would the G7/OECD Hiroshima AI Process Voluntary Reporting Framework (HAIP) provide us with a sufficient baseline? How would we get transparency on third-party providers?

3.  AI evidence for policymaking. It was agreed that since much of the research was in the private sector or for defense, there was no transparency on what was happening behind the scenes. Do we wait for peer review, or accept evidence from people’s experience and from papers published on sites like arXiv, an open-access repository? Where was the funding for research that was not technical but on AI impacts on health, education, environment, culture, and society?

4. Inclusiveness – It was clear that the Global South and other disciplines (besides AI) were being left out of conversations, and this was creating a skew in what values and metrics would be considered important for global governance of AI. For example, even with the recent The International Association for Safe & Ethical AI (IASEAI)elections, there is a conspicuous absence of the Global South. Though the 40-member Scientific Panel had strong representation of genders and the Global South and Global North, it was overwhelmingly composed of technical experts. The committee produced an excellent report - contratulations to Yoshua Bengio, Co-Chair, AI Scientific Panel and Maria Ressa, Co-Chair, AI Scientific Panel and the team!

Looking forward to the next Global Dialogue on AI Governance. Congratulations to the Co-Chairs of the UN Global Dialogue on AI Governance: H.E. Ambassador Egriselda López, El Salvador and H.E. Ambassador Rein Tammsaar, Estonia, and also to H.E. António Guterres, Secretary-General of the United Nations; Doreen Bogdan-Martin, Secretary-General, ITU; Khaled El-Enany, Director-General, UNESCO; Amandeep Singh-Gill, USG and Special Envoy for Digital and Emerging Technologies, ODET.