ComplyOn AI

Often overlooked but, success must include meaningful representation of fast-growing developing economies, including the Western Balkans. Countries like Albania, North Macedonia, and Serbia are not just policy followers. They are rapidly digitizing, experimenting, and in some cases moving faster than larger systems in adopting AI. If their perspectives are excluded, we risk creating frameworks that are misaligned with real adoption contexts. These regions face different constraints: limited regulatory capacity, high reliance on imported technologies, and strong incentives for innovation-led growth. A successful dialogue would therefore ensure: * Their inclusion in standard-setting conversations * Access to capacity-building and technical expertise * Flexibility in frameworks to avoid slowing down responsible innovation Finally, success would be a clear mechanism for continuity. Not a one-off event, but a structured process that tracks progress, supports countries at different stages, and evolves alongside the technology. If the Dialogue manages to balance global standards with local realities, especially by bringing emerging regions into the center of the conversation, it will have laid the foundation for governance that is not only principled, but actually workable.

• Safe, secure and trustworthy AI
• AI capacity-building
• Transparency, accountability, and human oversight

As founder of ComplyOn AI, we work with small and mid-sized businesses in the Western Balkans to help them understand and comply with AI rules like the EU AI Act. Based on that work, we chose these three priorities: Safe, secure and trustworthy AI: Most of our clients are just starting to think about AI risks. They need practical tools to assess whether the AI they use or build is safe and reliable. Without this foundation, no governance framework works in practice. AI capacity-building: In regions like the Western Balkans, awareness of AI governance is still very low. Businesses don't know what rules apply to them or how to follow them. Building that knowledge through training, simple guides and hands-on support is the most urgent need we see on the ground. Transparency, accountability, and human oversight: When AI makes decisions that affect people in hiring, lending, healthcare someone needs to be responsible. Our clients need clear, simple ways to document what their AI does and who is accountable when something goes wrong. This is the core of what good governance looks like in practice. These three areas are connected: you can't have accountable AI without the capacity to understand it, and you can't build trust without transparency. For emerging economies and smaller businesses, the global AI governance conversation needs to stay grounded in what is actually achievable not just what is ideal.

The compliance gap for small businesses: Most AI governance frameworks are designed with large corporations or governments in mind. Small and mid-sized businesses, which make up the majority of economies in the Global South and emerging markets, often lack the legal teams, technical staff, or budgets to implement what is being asked of them. If global AI governance ignores this reality, the rules will exist on paper but not in practice. Geographic and regulatory fragmentation: Countries and regions are developing AI rules at very different speeds and in very different directions. For businesses operating across borders especially in regions like the Western Balkans, which are EU-adjacent but not yet members, this creates serious confusion. There is no clear guidance on which rules apply, how they interact, or what happens when they conflict. This needs dedicated attention, not just a footnote under "interoperability."

The Western Balkans sits in a difficult position. The region is economically tied to the EU and is expected to eventually adopt EU rules including the AI Act but most countries are not yet members and have no formal seat at the table where those rules are being made. This means businesses and governments in the region are expected to comply with a regulatory framework they had no voice in shaping. At the same time, local AI governance capacity is almost nonexistent. There are very few lawyers, consultants, or public officials who understand what the EU AI Act requires in practice. Most small businesses using AI tools whether for customer service, hiring, or finance have no idea they may already be subject to compliance obligations. The awareness gap is significant. The region is also at an interesting inflection point. Because AI governance infrastructure is still being built, there is a real chance to do it right from the start rather than retrofitting rules onto systems already in place. Countries that move early and build credible frameworks will have a competitive advantage in attracting EU-aligned investment and digital business. There is also strong appetite among young professionals and emerging tech companies in the region to engage with these issues. Civil society, startups, and academic institutions are increasingly interested they just need structured support and access to resources.

The AI Dialogue has a unique opportunity to do something that most existing forums have not managed: bring the full diversity of the world's experience with AI into one conversation, and turn that into concrete, shared action. Most international AI governance discussions have been dominated by a small number of powerful countries and large technology companies. The result is frameworks that reflect their priorities, their infrastructure and their resources. The Dialogue can change this but only if it is designed to genuinely include voices from smaller economies, emerging markets, and regions that are often consulted last and listened to least. Concretely, the Dialogue can play several important roles: Connecting rule-makers with practitioners: There is a large gap between the people writing governance frameworks and the businesses and citizens living under them. The Dialogue can serve as a bridge surfacing what is actually working on the ground and feeding that back into policy design. Supporting capacity in underserved regions: Not every country can build an AI regulatory body from scratch on its own. The Dialogue can help coordinate technical assistance, share model frameworks and match countries that need support with those that can provide it.

The EU AI Act is the most comprehensive binding AI regulation to date and a useful reference point globally. The OECD AI Principles and UNESCO's AI Ethics Recommendation provide a broadly endorsed values foundation. The Dialogue's added value is its universal reach : it can connect these efforts, identify where they conflict or overlap, and help build bridges between them. It can also go further than existing forums by centering implementation not just principles. The gap today is not a lack of values statements. It is the absence of practical, accessible tools that help governments and businesses actually apply those values in the real world. That is where the Dialogue can make a lasting difference.

The Dialogue should not be a single annual event. It needs a permanent structure with clear working groups, regular reporting, and transparent decision-making. Key elements should include: A multi-stakeholder advisory body that feeds into official deliberations, with rotating representation to prevent capture by any one group. Regional hubs that allow local conversations to feed into the global process so that the Western Balkans, Sub-Saharan Africa, or Southeast Asia are not always reacting to agendas set elsewhere. Open consultation rounds where any organization or individual can submit input in accessible formats, not just formal written submissions requiring legal expertise. Plain-language outputs every major agreement or recommendation should be translated into practical guidance that a small business or local government can actually use. The goal is a Dialogue that looks like the world it is trying to govern.

The gap between who is making AI governance decisions and who is most affected by them is wide. Several groups are consistently left out. Who is missing: Small and mid-sized businesses are the backbone of most economies and the primary deployers of AI in everyday contexts yet governance conversations are almost entirely shaped by large corporations. Their practical constraints and experiences are rarely reflected in the frameworks being built. Emerging and transition economies including the Western Balkans, Central Asia, Sub-Saharan Africa, and parts of Latin America and Southeast Asia are expected to adopt governance standards they had little role in designing. Their regulatory realities, resource limitations, and development priorities are different, and those differences matter.

Most international governance forums follow the same format: prepared statements, panel discussions, and closed negotiating rooms. That approach produces documents, not dialogue. The AI Dialogue needs to do something different. Formats worth trying: Problem-based working sessions instead of organizing discussions around themes, organize them around specific, concrete problems. "How should a small business assess whether its hiring tool is discriminatory?" produces more actionable output than "ethics of AI in employment." Cross-regional peer exchange structured sessions where practitioners from different regions share what has actually worked and what has not, without the pressure of formal negotiation. Honest peer learning is often more useful than polished presentations.

From COA's work with businesses in the Western Balkans, the most effective approaches share common features: they are written in plain language, they start with risk assessment rather than compliance checklists, they are proportionate to the size and capacity of the organization, and they are updated regularly as technology evolves. Governance that is too complex to implement is governance that will be ignored. The best examples in the world are the ones that make compliance accessible, not just technically correct.