„Източник: Bulgaria On Air“

Emiliyan Arnaudov commented in the studio of Bulgaria Morning on Bulgaria On Air TV what guarantees the EU AI Act- the world’s first comprehensive AI law, which came into force on 1 August.

The main focus of the Act is to protect democratic principles, the rule of law and human rights. These are the highlights set out in the Act which gives three types of regulation. On the one hand, it is the definition of the various risks accompanying AI and the risk-management measures. In view of what the relevant system is set to do, an AI system is defined as High-risk when it has the potential to damage public relations. Others are qualified as General-purpose AI systems such as chatbots with AI, already familiar and used by a very large number of people, which raises the need for transparency in their functioning.

Тhe Act sets out certain prohibitions on AI in order to avoid abuses related to a person being identified when he or she does not want or need to be in a publicly accessible space (e.g. applications related to facial recognition based on a randomly collected broad database of images of people on the Internet). Att. Arnaudov explained that this hypothesis differs from the facial recognition of our phones, where we want to be recognized but we do not want the image that we have given to allow us to be recognized when we enter a restaurant, for example. Some exceptions to the prohibition on the use of facial recognition in public places are set. Such can only be carried out in three strictly defined hypotheses and only with specific authorization by a judicial authority:

• if a suspect is wanted for a serious crime;
• if a victim of a crime is wanted (kidnapped person, victim of trafficking, etc.)
• where there is evidence of an imminent terrorist attack;

The Regulation prohibits the uses of AI to attempt to qualify people on the basis of biometric data by making assumptions about their political and religious views, their sexual orientation, as well as attempting to identify their ethnic or racial characteristics. Attempts to classify people as potentially likely to commit a crime on the basis of AI are also prohibited.

The Regulation focuses mainly on how AI is used by public administrations on the one hand and companies that develop and use AI on the other. One of the important things that the Regulation does is to define when an automated system can now be considered artificial intelligence. It is important to note that AI is present when there is some independence from human intervention and the AI system itself makes independent inferences, algorithms, etc.

In the use of AI, there are generally two types of problems that can arise at the intellectual property level – at the “input” and at the “output” of the system. In the case of the “input” of the system, things are largely regulated by Directive (EU) 2019/790 Of the European Parliament and Of the Council on copyright and related rights in the Digital Single Market, which was transposed in Bulgaria with effect from 1 December 2023 and according to which the free collection of data, regardless of the protection of intellectual property on the objects concerned, may only be carried out when it is for strictly scientific purposes. Where data collection is for the purpose of some commercial exploitation, then rights holders have the right to appropriately prohibit their subject matter from being used during “input” into the AI. Аccording to Att. Arnaudov the second group of problems has a greater potential for adverse consequences to arise is in the “output” of the system, i.e. how the issue of intellectual property over something that is created by AI is regulated. Based on the whole theory of copyright over the years the copyright holder can only be a human being, therefore, what is created by AI is not protected by copyright and related rights. According to Att. Arnaudov, however, the practical problem that may arise in markets such as Bulgaria, where not a small part of businesses does not want to pay to settle their copyrights, is that once objects created by AI are not protected, they can be used for free. That means AI may create for the audience which does not want to pay for a creative product a free alternative which in the long run would lead to damage to the real creative sector.

One of the EU’s objectives in this Regulation is to promote innovation and competitiveness based on AI. The Regulation stipulates that by 1 August 2026, each EU Member State must set up a National AI Lab to assist start-up SMEs in experimenting in a closed and secure environment with the benefits and developments they could have, using AI to improve their competitiveness.

This was yet another participation in which colleagues from the Law Firm commented the topic. In the spring Att. Koparanov was a guest on Bulgaria ON AIR on the case of the distribution of fake pornographic photos generated by AI.