---

Twenty eight countries gathered in Bletchley Park , England in November 2023 to discuss how safe is AI. This elite gathering, consisting a handful of lab CEOs and surprisingly, USA and China in the same room, worried that frontier AI models could help someone synthesise a bioweapon, or that humanity was sleepwalking into a loss-of-control scenario. The Bletchley Declaration, which was non-binding set out that AI should be “safe, human-centric, trustworthy and responsible.”

Six months later, the AI Seoul Summit May 2024, brought together representatives from 27 countries, and the European Union to sign the final “Seoul Ministerial Statement and while the Seoul Declaration which is a leaders’ level commitment was endorsed by 10 countries and the EU. Safety rightfully, remained central. The discussion also focused on “innovation” and “inclusivity” at the summit. The Seoul Declaration explicitly referenced the UN Sustainable Development Goals and digital divides. The Seoul Declaration was a shift from treating AI safety as a just national concern toward to more of a global scientific and governance project. As for the Inclusivity, coupled with sustainabe eco-systems, received explicit attention in the Ministerial Statement however on the environmental impact.

Paris in February 2025 AI Summit got a recast. It moved from being called an AI Safety Summit to AI Action Summit. The focus from“how do we prevent catastrophe” moved to “how do we capture the opportunity”. The summit covered five main topics: 1. public interest, 2. future of work, 3. innovation and culture, 4. trust in AI, and 5. global governance. The official document was the “Statement on Inclusive and Sustainable AI for People and the Planet” was signed by 62 entities (58 on the day, later others). This included a broad coalition of nations like France, India, China, Japan, Canada, Brazil, and Australia, as well as the European Union and the African Union Commission. Sending the world in to geopolitical frenzy, UK argued the text was too vague on national security specifics and USA felt the language was too regulatory and “precautionary,” which they feared would hand an advantage to China. They both did not sign the declaration. Although inclusivity was a major pillar, the framing came at the expense of safety. The summit shifted the conversation from governing frontier AI risks to expanding AI access and investment.

February 2026 the AI Impact Summit was held in India. Did it finally hit the mark?

India 2026: The First Inclusive Summit?

New Delhi’s AI Impact is different in one important way. It’s the first summit where the host country’s national context forced the conversation toward deployment realities rather than frontier development. The Global South set the agenda. The summit moved the focus from “preventing AI apocalypse” to “spreading AI opportunity. Eighty eight to 89 countries. Including the US, UK, andChina, signed the New Delhi Declaration on AI Impact.The Times of India says it was one of the largest tech gathering in history, with over 1,000 global delegates and 500,000 visitors to the exhibitions. Photos of hand holding, and lack thereof, Chinese-made robot dog presented as indigenous innovation, to protests inside the venue made headlines.

The declaration was structured around seven pillars: Democratising Resources, Economic Growth, Trusted AI, AI for Science, Social Empowerment, Human Capital (Skilling), and Energy Efficiency. But for all its ambition, the gap between the rhetoric of “welfare for all” and the actual governance architecture being built is real and worth naming.

What Four Summits Mean for SMEs?

Here are the questions that deserve a place on the stage at these gatherings: Why is AI “adoption” as “meaningful structural transformation, for SMEs in most emerging markets? what are the actual structural barriers, and what should government policy look like to address them?

SMEs represent 90% of formalized businesses globally and employ the majority of the workforce in virtually every emerging economy. The AI governance conversation has spent three years designing frameworks for the frontier model layer and the geopolitical layer. They had the inclusivity as a topic, but never specifically mentioned that SMEs, should be part of this topic.

Eurostat reports that in 2025, only 17 percent of small enterprises, and 30.36 percent of medium enterprises used AI in their businesses. At large enterprise this was 55.03 percent. When it comes to the definition of usage, it was mere use. Not actual digital transformation Even so, this difference might be explained for simple reasons. The report highlights that, complexity of implementing AI technologies in an enterprise, financial stability, regulation barriers and costs could explain the difference. In an OECD survey covering four G7 countries, 50 percent of SMEs report that their employees lack the skills to use generative AI. Connectivity gaps, lack of AI-ready data and infrastructure, low AI literacy and weak business strategy, skills shortages, limited access to capital, and regulatory uncertainty are given as the barriers

My attempt to get insights in to SME adoption, at least mere AI usage came as unavailable. Perhaps that is a whole different research for a later date.

However I feel that pattern is the same or more even in the emerging markets. Having worked in the implementation side of technology adoption in emerging markets I can confidently say that I have seen it all before. The problem with SME technology adoption has never been the technology. It’s institutional and it is about basic economies of scale. The absence of financing instruments to adapt to small-business risk, the lack of training, and most importantly the regulatory uncertainty that makes SMEs to shy away from meaningful AI adoption. These are solvable policy problems. They require deliberate attention from the people in the room at these summits.

What I Saw In the Leaders’ Declaration This Year?

I am seeing the leaders declaration getting scrutinised. This is a good sign. The New Delhi Declaration has been signed by 91 countries (this is an impressive number and importantly includes UK and USA), have indirectly mentioned some benefit for SMEs. But then again they’re all voluntary, non-binding, and several layers removed from the actual adoption barriers for SMEs. There is a charter for Democratic Diffusion of AI; a framework to promote affordable access to foundational AI resources and support locally relevant innovation, The Global AI Impact Commons; a platform to share and replicate successful AI use cases across regions and, The Frontier AI Impact Commitments; companies agreed to publish anonymised data on how their AI systems are being used which could, over time, give policymakers actual evidence on SME adoption patterns and re-do/create relevant policies. They are just promises as of now, and in a sense, just feel-good items.

Added to all this in the same summit White House official Michael Kratsios explicitly stated at the summit: “We totally reject global governance of AI” . They offered “trade over aid” as America’s model for AI diffusion. Essentially AI will be a product that requires investments that SMEs are struggling to get.

Lets hope we will hear more of SMEs in  Switzerland, who is scheduled as host the next major global AI summit in Geneva 2027.

This post is part of my ongoing research on AI policy barriers and SME adoption in emerging markets. More will be said in my next blog posts.