From research to a practical question
The idea began while I was writing my MBA thesis on explainability under the EU AI Act. To understand the broader context, I explored different approaches to governing AI: the EU’s comprehensive, horizontal framework; China’s more vertical approach, with rules targeting particular technologies and services; and debates in the United States about new legislation and the adequacy of existing legal protections. I also examined ethical frameworks developed in countries such as India, Saudi Arabia and the UAE, alongside standards and guidance from organizations including NIST, IEEE and ISO. Despite differences in their scope and legal status, these approaches repeatedly addressed similar concerns: privacy, fairness, reliability, transparency and human responsibility.
Throughout this research, one debate was impossible to avoid: would regulation encourage AI adoption by building trust, or hinder it by increasing costs and constraining innovation? My research suggested that European regulation sometimes carries too much of the blame for Europe’s wider difficulties with competitiveness and productivity. Nonetheless, translating broad legal requirements into workable technical and organizational practices presents considerable challenges. Comprehensive legislation on a technology developing this quickly is a tall order, which also helps explain the important role given to technical standards in implementing the AI Act.
This made me consider what approach North Macedonia should take. European integration makes the EU framework particularly relevant, but I remain cautious about treating additional binding legislation as the immediate answer. I am much more convinced that businesses need practical guidance on responsible use, whatever legislative approach we eventually adopt.
What adoption looks like in practice
These questions were becoming increasingly concrete in my own law office. We use AI to assist with research, draft documents and court submissions, correspondence, service offers, and business strategy. Its value extends beyond producing text faster: it can help organize complex information, explore alternative arguments and reduce repetitive work, leaving more time for analysis and clients. These experiences have made me enthusiastic about its potential, particularly for smaller businesses whose capacity is often constrained by time and resources.
They have also exposed limitations that are easy to overlook. Working in Macedonian can reveal weaknesses that are less apparent in English. Incomplete digitalization means that relevant records, documents and professional materials may be unavailable to a system, even when its answer sounds comprehensive. I have also become concerned about overreliance and professional competence. If younger colleagues routinely receive finished answers without learning to research, evaluate sources and develop their own reasoning, we risk weakening the expertise they will eventually need to supervise the technology.
At the same time, I have encountered excellent professionals who hesitate to use AI precisely because they take their responsibilities seriously. Their concerns about confidentiality, accuracy and the interests of their clients are reasonable. Yet without practical guidance, they may avoid even applications whose risks could be managed. More enthusiastic adopters face a different difficulty: they may discover useful applications before recognizing the implications of the information they provide or the outputs they rely on.
Both groups can consist of conscientious people trying to do the right thing. Leaving them without a clear basis for making decisions can produce either excessive hesitation or insufficiently examined risk-taking. The first can mean missed productivity gains and fewer opportunities to serve clients; the second can cause harm, undermine trust and expose a business to reputational damage and liability.
From a set of principles to a wider initiative
My initial idea was to prepare a set of responsible AI principles that could help businesses navigate these questions. Other countries, including Serbia, had already demonstrated the usefulness of voluntary ethical guidance. North Macedonia already had valuable AI initiatives underway, including Vezilka and discussions organized through Metamorphosis’s e-Society conference. What we saw missing was a comprehensive set of responsible AI principles that businesses could use as a common reference across different applications and sectors. We wanted to help fill that gap by bringing established international guidance together in an accessible document, adapted to the circumstances and everyday decisions of businesses in North Macedonia.
Angela Angjelovska saw the potential for something broader. Drawing on her experience in civil society and organizational work, she proposed developing an initiative that could bring people together, support research and sustain the discussion beyond a single document. That helped turn my original idea of writing principles into our shared project, Ethical AI Macedonia. The document would provide a starting point, while the initiative would create opportunities to test its usefulness, hear different perspectives and learn from businesses’ actual experiences.
In drafting the principles, I drew on the more developed standards, frameworks and guidance I had studied during my thesis research. I brought those ideas together and added emphasis where my own experience suggested that our local circumstances deserved particular attention: smaller languages, incomplete digitalization and the effects of AI on professional competence. We are not presenting their ethical foundations as groundbreaking. Their intended value lies in making established ideas accessible, connecting them with our circumstances and giving businesses a useful basis for discussion.
What we want to achieve
This first draft is open to review, comments and improvement. We want businesses, technical practitioners, researchers and other interested people to challenge it, identify omissions and help us understand which recommendations are useful or difficult to apply. Their experience should shape both the document and our future work.
More broadly, we want to gather businesses around these issues, learn about the challenges they face and exchange practical approaches to addressing them. The principles, company research and discussions such as our November event are the first steps in that process. Our purpose is to encourage greater AI adoption through responsible use: helping businesses benefit from the technology while protecting the people they serve, their own interests and the professional capabilities on which their future depends.