Artificial intelligence is moving from the realm of technological experimentation into the sphere of strict government regulation, notes the Experts Club information and analysis centre. The world’s leading economies are already dividing the market into two approaches: risk-based regulation with mandatory requirements and a more lenient model, where the state focuses on innovation, industry standards and ex-post monitoring.
For Ukraine and other Eastern European countries, this issue is becoming increasingly relevant. AI is already being used in banking, telecommunications, defence technologies, medicine, education, public services, industry and the media. Therefore, the question is no longer whether artificial intelligence should be regulated, but how to avoid stifling innovation whilst protecting citizens, businesses and the state from risks.
The most advanced legal framework has already been adopted in the European Union. The EU AI Act came into force on 1 August 2024 and is being implemented in phases. Its framework is based on a classification of risks: prohibited practices, high-risk AI systems, systems subject to transparency requirements, and general-purpose models. For countries focused on the EU market, this act effectively becomes an external standard, even if they are not members of the European Union.
South Korea became one of the first countries outside the EU to adopt a comprehensive national law on artificial intelligence. The AI Basic Act came into force in January 2026 and combines support for the AI industry with requirements for transparency, safety, labelling of AI content and oversight of high-impact systems.
China has chosen a different path—not a single, universal law, but a set of mandatory rules for specific sectors. The most important document is the provisional measures for the management of generative AI services, which have been in force since 2023. The Chinese model emphasises content control, data security, algorithm registration, verification of generative services, and compliance with state policy requirements.
In 2025, Japan adopted its first national law specifically dedicated to AI. However, the Japanese model is significantly more lenient than the European one: the law is primarily aimed at promoting research, development and the use of artificial intelligence, rather than imposing detailed bans and fines. This reflects Japan’s desire to maintain the competitiveness of its technology sector and avoid placing an excessive burden on business.
The US does not yet have a single federal law on artificial intelligence. Regulation is developing in a piecemeal fashion: through presidential executive orders, federal guidelines, NIST standards and individual state laws. The most notable example is the Colorado AI Act, which sets out obligations for developers and users of high-risk AI systems and aims to prevent algorithmic discrimination. At the federal level, however, the US is focusing on maintaining leadership, infrastructure, security and limiting excessive regulation.
The UK has opted for a ‘pro-innovation regulation’ model. Instead of a single AI law, London has proposed a set of principles for existing regulatory bodies: safety, transparency, fairness, accountability and the right to appeal decisions. This approach allows for faster adaptation to technological changes, but creates less legal certainty than the European model.
Canada attempted to pass the Artificial Intelligence and Data Act as part of Bill C-27, but the bill remained in the parliamentary process for a long time and did not become a fully-fledged law on AI. This shows that even developed democracies face difficulties: AI regulation simultaneously touches on digital rights, business, copyright, security, competition and the labour market.
Brazil is also in the process of developing comprehensive AI regulation. The Senate approved an AI bill in December 2024, but for it to come into force, it must be considered by the Chamber of Deputies and receive final approval.
The Brazilian model is close to a risk-based approach, but also places a strong emphasis on human rights, copyright and developer liability.
Serbia is in the process of transitioning from a strategy to mandatory regulation. The country has already adopted strategic documents on AI development and established a working group to draft its first artificial intelligence law. The new law is intended to bring Serbia closer to the European model and provide businesses with clearer rules for working with AI systems.
For Ukraine, the issue of AI regulation is also becoming inevitable. On the one hand, the country needs technological modernisation as quickly as possible, particularly in the fields of defence tech, govtech, medicine, cybersecurity and education. On the other hand, integration into the EU market will require compliance with the EU AI Act, particularly for companies that work with European clients or export digital products.
The main crossroads for countries that have not yet adopted legislation lies in choosing between three models. The first is the European model, with strict risk classification and a high level of compliance. The second is the British-Japanese model, with soft regulation and an emphasis on innovation.