How is Serbia Responding to the Rapid Growth of Artificial Intelligence?
Bojana Račanović, who works with PwC as an Attorney at Law, explains the extent to which Serbia is prepared for the risks that accompany the rapid expansion of AI across the globe.
Artificial intelligence (AI) may be defined as software that is trained by a particular set of data to perform specific tasks (eg, recognition of specific patterns). Numerous activities that require learning, decision-making and other forms of human intelligence can also be performed by AI systems. These activities range from simple tasks such as proofreading, translation, or facial recognition to more complex tasks requiring advanced algorithms that could, for example, diagnose cancer, drive a car or even approve a bank loan.
AI has characteristics of a general-purpose technology, as its application spreads to all areas of the economy and society – with the potential to bring about revolutionary changes in many of them. Globally, AI will lead to significant changes in everyday life, work and business. Its presence is already visible at an economic and societal level and will only develop more significantly in all other aspects of human life in the near future.
However, along with the undeniable benefits introduced by AI – especially in terms of economic development and efficiency improvement – come certain challenges. At this point in time, it is crucial to address the challenges and risks linked to AI to ensure ethics and safety kapply in the process of AI development, especially with regard to:
- the protection of personal data;
- the prevention of discrimination when AI is used for decision-making purposes; and
- the establishment of responsible AI.
In this respect, all countries should regulate AI within their jurisdiction and in accordance with the highest ethical and safety standards.
How Is AI Regulated in Serbia?
Serbia has already enacted strategic documents and guidelines to regulate the development, implementation and use of AI. As a candidate for EU membership and a participant in the European Union Framework Programme for Research and Innovation, Serbia aims for its regulations to comply with those of the EU wherever necessary when it comes to AI.
In 2019, Serbia adopted the Strategy for the Development of AI for the period 2020–2025, which determined the country’s objectives and measures for the further development and application of AI in different fields.
“Serbia aims for its regulations to comply with those of the EU wherever necessary when it comes to AI.”
In March 2023, the Serbian government adopted Ethical Guidelines for the Development, Implementation and Use of Reliable and Responsible AI (“the Guidelines”), which were harmonised with the Proposal for a Regulation on AI announced by the EC in 2021. Given that representatives of Serbia participated in the creation of the UNESCO Recommendation on the Ethics of AI in 2021, some of these recommendations are implemented in the Guidelines.
The aim of the Guidelines was to enable Serbia to adopt responsible AI that would benefit society in such a way that the latest capabilities of AI are utilised while all associated risks are manageable. Although the Guidelines are not binding in nature, they set out the following general principles for the creation of reliable and responsible AI that all individuals and legal entities developing and using the AI systems should observe.
- Explainability and verifiability – this emphasises transparency and the verification of the AI system throughout its entire life cycle.
- Dignity – this means ensuring the personality, freedom and autonomy of humans to the extent that they can never be subordinated by the AI system.
- The principle of “do no harm” – this means the AI system must comply with safety standards, must contain mechanisms for avoiding damage to people and their property, and must not be used for malicious purposes.
- Fairness – this protects the rights and integrity of people, especially those in sensitive categories (eg, people with disabilities).
The conditions for the creation of reliable and responsible AI are based on the above-mentioned general principles and include:
- action (mediation, control, participation) and supervision;
- technical reliability and safety;
- privacy, data protection and data management;
- transparency;
- diversity, non-discrimination and equality;
- social and environmental well-being; and
- responsibility.
- For each condition, the Guidelines
- introduced a self-assessment questionnaire and
- recommendations for complying with the stipulated conditions.
The aim of the questionnaire is to enable individuals and/or legal entities that develop, apply and/or use AI systems to assess their compliance with the above-mentioned conditions and then identify areas in which improvements are necessary.
In addition to the specific AI-related regulations, the Serbian Personal Data Protection Act applies in respect of the personal data fed into AI systems. Also, legal entities in Serbia may be bound by the EU Artificial Intelligence Act (the “AI Act”) – a draft of which was approved by the EU Parliament in June 2023. When the AI Act is adopted, it will impose a set of obligations both on providers and deployers of in-scope AI systems used (or producing effects) in the EU, irrespective of their place of establishment.
Which AI Systems Are Considered High-Risk in Serbia?
There are numerous risks associated with AI systems and Serbia has determined several high-risk AI systems. Pursuant to the Guidelines, high-risk AI systems have been identified as those that are used in the following areas:
- biometric identification and categorisation of individuals;
- management of critical infrastructures such as transport, gas, electricity – ie, those in which the life and health of citizens could be put at risk;
- educational or vocational training that may determine the professional course of someone’s life or access to further education (eg, scoring of exams);
- employment, management of employees/hired persons, and access to self-employment (eg, CV-sorting software for recruitment procedures);
- healthcare (specifically, analysis of genetic and health data);
- access to and use of public social services as well as essential private services (eg, credit ratings that might deny citizens the opportunity to obtain a loan);
- criminal prosecution – in particular, using personal traits, characteristics, or previous criminal behaviour to assess the risk of an individual committing or re-committing criminal offences;
- migration, asylum and border control management (eg, verification of authenticity of travel documents); and
- administration of justice and democratic processes (eg, applying the law to a concrete set of facts).
This list is used as a guide to representative examples of high-risk AI systems and will be updated accordingly.
High-risk AI systems are not considered undesirable in the market. However, given the importance of the areas of life in which they are applied and the potential range of impacts on an individual and his/her integrity, it is necessary to analyse such AI systems individually and assess their associated risks.
Who Is Accountable for an AI System in Serbia?
In accordance with the Guidelines, designers and developers bear a high level of responsibility for the design review, development, decision-making and results of the AI system. However, as AI systems are usually created in a multidisciplinary team, all individuals participating in the creation of the AI system at any stage of the process are accountable for its impact to a certain extent. This includes companies that have invested in its development.
Laws regulating the aforementioned responsibilities of individuals and legal entities in further detail have yet to be enacted in Serbia. They are anticipated, nonetheless.
PwC Legal
