In light of the rapid development of artificial intelligence (AI), there is growing concern about creating the right conditions for the ethical and responsible use of the latest technologies. In response to this problem, the world's leading countries, such as the United States, Britain, and the EU, are creating their own regulatory frameworks for the development and use of AI. 

Ukraine, which already uses AI systems in various industries, also supports initiatives to regulate these technologies. This year, in June, the Ministry of Digital Transformation presented a White Paper outlining the future strategy of Ukrainian regulation of artificial intelligence. In this context, it is necessary to determine whether the Ukrainian approach will be sufficient to achieve the goals outlined in the White Paper.

Ukrainian approach to AI regulation

Last year, the Ukrainian government considered various strategies to implement artificial intelligence regulation in Ukraine. The top priority was a strategy based on an inductive approach, which meant gradual implementation of regulation as AI technologies evolved. 

Subsequently, in October 2023, the Ministry of Digital Transformation published the Roadmap for the Regulation of Artificial Intelligence in Ukraine, which for the first time defined the Ukrainian approach to technology regulation, aiming to "find a balance between business interests and ensuring an adequate level of protection for citizens." Taking into account the gradual movement toward the implementation of the European AI legislation, the Ministry of Digital Transformation chose a "bottom-up" approach in order to "not overregulate the AI market" at the initial stage of regulation development.

"Bottom-up" is one of the modern approaches to regulating artificial intelligence that is currently being used, in particular, in the UK. The main feature of the approach is to promote "self-regulation" of the AI industry, which involves providing tools and recommendations for businesses and other stakeholders on the development and use of AI technologies. Despite the adaptability of the approach and its openness to innovation, the use of bottom-up tools may be insufficient to protect against the risks that may arise in the course of AI application.

The Ukrainian version of the regulation was described in more detail in the White Paper on AI Regulation in Ukraine published on June 26, 2024. 

According to the Kantar Ukraine survey presented in the introduction to the White Paper, the vast majority of Ukrainians know what AI technologies are. A significant proportion of citizens also have experience using AI tools. Among criminal justice lawyers, the results of this year's sociological survey by the Dnistrianskyi Center and Fama Agency also showthat there is a high level of awareness of AI and a high level of use of these technologies. 

Due to the significant popularity of AI technologies in Ukraine, the Ministry of Digital Transformation recognized the need to develop its own regulatory strategy, reapproving the use of the "bottom-up" approach. In this regard, it can be argued that the Ministry of Digital Transformation's vision of AI regulation in Ukraine has not changed.

The main goals of the approach were identified:

• support for the competitiveness of the AI industry, which involves providing tools for Ukrainian businesses to enter international markets;
• protecting citizens' rights from potential risks and malicious use of AI;
• European integration, which in the future would mean the implementation of the EU AI Act in Ukraine.

The regulation will be implemented in two stages. At the first stage, the Ministry of Digital Transformation plans to provide "extra-legislative" tools to help businesses and the state prepare for future AI regulation. For example, the key tools proposed by the Ministry are the methodology for evaluating AI systems, the regulatory sandbox, and the labeling of AI systems. 

Artificial intelligence will be evaluated to determine which products will fall into the regulatory sandbox and to identify the risk level of a particular system. The methodology will be based on a risk-based approach that will be used in the framework of EU AI legislation. The methodology will be adapted to the Ukrainian context based on current EU and Council of Europe practices. 

The regulatory sandbox will be an environment where AI developers can test their systems to meet future regulatory requirements. According to the Ministry of Digital Transformation, the regulatory sandbox will only include products that have a "medium or high human rights impact" or "meet other selection criteria." For those products that do not participate in the regulatory sandbox, a legal assistance platform will be offered to help them comply with future legislation.

In addition, the Ministry of Digital Transformation will encourage artificial intelligence developers to voluntarily label their products, which involves providing information about the construction, training data, system functions, etc. The purpose of such labeling is to increase transparency and cover the requirements set forth in EU legislation. The level of disclosure of AI systems is voluntary and is decided primarily by AI companies.

Other tools offered by the state include recommendations on the use of AI for various industries, codes of conduct for businesses, and the Center for Responsible AI web portal. After the implementation of the instruments, the Ministry of Digital Transformation expects a "comprehensive and multilateral positive" impact that will prepare the state, citizens, and businesses for the second stage of regulation.

The second stage envisages the gradual implementation of the EU AI Law. The Ministry of Digital Transformation plans to start drafting a law on the use of AI after the EU law is finally adopted. At the same time, the authors of the White Paper point to the prospect of maintaining the "bottom-up" approach during the implementation phase, arguing that it is possible to postpone the entry into force of certain provisions of the EU Law. It is important to note that the Ministry of Digital Transformation does not plan to adopt provisions that contradict EU legislation or to mix and match approaches from other countries while preparing the draft law.

What should you pay attention to?

The White Paper on AI Regulation in Ukraine outlines a bottom-up approach to AI regulation in Ukraine, which envisages "self-regulation" of AI using new tools for the state and business, ensuring a certain level of flexibility and adaptability to innovations. Therefore, if properly implemented, the proposed measures can become an important step for the development of the AI industry in Ukraine. In this context, it is also important to consolidate the course of European integration and move toward the implementation of EU legislation on AI.

At the same time, the Ministry of Digital Transformation identifies the lack of financial and organizational resources for "rapid implementation of mandatory regulation" in times of war as one of the limitations of the approach. However, when choosing a "bottom-up" approach in the field of AI, it is necessary to take into account not only material resources but also the legal environment of the state. "Bottom-up" does not mean a complete rejection of legal regulation. To implement the approach, it is necessary to rely on existing legal frameworks, such as laws regulating related areas, which the authors of the White Paper also recognize.

The previous remarks on the bottom-up approach in the Ukrainian context are not only about doubts about ensuring proper human rights protection, but also about the relevance of current Ukrainian legislation to the development and use of AI. At the current stage of regulation, in order to properly implement the approach and provide guarantees for the protection of human rights in the use of AI, it is first of all necessary to revise Ukrainian legislation in the digital sphere, which, according to the expert community, is not adapted to AI technologies.

In addition, the authors of the White Paper recognize that there is no alternative to adapting EU legislation on AI and the digital sphere in Ukraine in the future, while noting that it is possible to maintain a "bottom-up" approach at the implementation stage. The start of the implementation of the EU Artificial Intelligence Law, regardless of the gradual adoption of certain provisions, will mean the abandonment of the bottom-up approach, as the EU AI Law provides for strict regulation and a top-down approach to regulation.

The tools provided for at the current stage of regulation also need to be better specified.

The White Paper suggests that only a certain part of AI products will be of interest for evaluation and testing within the regulatory sandbox. In particular, it refers to technologies that have a "medium or high impact on human rights." At the same time, the process of selecting technologies that will fall into the "medium" or "high" impact categories remains unclear. 

The issue of competent model evaluation is extremely important, as the lack of proper measurement of the impact of AI systems remains one of the global challenges for technology researchers. The authors of the annual AI Index report by Stanford University emphasize that the absence of a standardized AI assessment complicates the process of comparing the impact and risks of different AI models. Thus, the absence of a clear methodology for assessing AI makes it impossible to make informed decisions on the classification and selection of AI technologies. This emphasizes the need to develop appropriate criteria for selecting and assessing the impact of AI systems on human rights and society.

Also, as noted earlier, the developers of the approach envisage the possibility of voluntary labeling of AI, which, according to the White Paper, means "clear and structured information" about the AI system. However, due to the voluntary nature of labeling, the boundaries of disclosure of such information are rather blurred. The overwhelming voluntariness, while providing a certain level of flexibility for businesses, does not provide clear standards for what system labeling should look like, particularly in the context of preparing for the future implementation of EU AI legislation.

Conclusions and recommendations

The White Paper was the first document to define in detail the strategy for regulating AI in Ukraine. Although the Ukrainian approach is promising in the context of artificial intelligence development, it still needs to be improved in terms of protecting the interests of society and preparing for the implementation of the EU AI Law.

The following recommendations should be taken into account to ensure the quality implementation of the current stage outlined in the White Paper:

• Improve the current Ukrainian legislation on the development, training, and operation of AI models. The use of bottom-up tools, even without the development of special laws on artificial intelligence, requires legitimate grounds for the development of the AI industry. It is important to identify and minimize gaps in Ukrainian legislation regarding the development and deployment of AI technologies, especially in laws regulating related areas, such as the Law of Ukraine "On Personal Data Protection." Such changes will help ensure the proper level of protection of citizens' rights when using AI technologies at the current stage of regulation.
• Define clear criteria for evaluating artificial intelligence. Ukrainian regulators should establish categories to determine the risk levels of AI systems. This necessitates a clearly defined methodology with a set of tools and indicators for calculating the impact of a particular system. In addition, the methodology should be flexible enough to ensure that it is constantly reviewed and improved.
• Develop standards for labeling AI systems. Clear guidelines for determining what system labeling should look like are especially important for companies planning to enter the EU market. The standards for providing information about AI products should be close to the provisions of the EU AI Law, which provide clear requirements for technical documentation to be provided by AI developers. 

In order to ensure high-quality preparation for the second stage, namely the implementation of EU AI legislation, it is necessary that current tools and approaches bring Ukraine as close as possible to future regulation. The tools proposed by Ukrainian regulators should not only increase the capacity of the state and business, but also provide guarantees for the protection of human rights, while meeting future regulatory requirements in the context of European integration.