As Artificial Intelligence (AI) becomesan important part of our lives it needs to be regulated. The EU has taken on the task and has accepted the AI Act. But does the AI Act regulate the best or are there any other way? What impacts does the AI Act have on legal technology innovation? In this blogpost I attempt to answer these questions.

What is AI, and why should it be regulated?

One of the most hotly debated topics of our time, and one of the terms we use most often is AI. Although the public started to pay attention to the topic with the 2022 release of ChatGPT, developed by OpenAI, the technology had been around long before that. Artificial intelligence refers to tools that work in a way that is as close as possible to human intelligence, approximating the functioning of the human brain depending on their level of sophistication. The development of AI began in the 1950s, when efforts were made to create machines that could pass the test that Alan Turing had designed. According to the Turing test, AI is only truly functional if a human opponent cannot tell the difference between a human and a machine system. Although there have been several boom periods in AI developments over the last century, these have usually been followed by a downturn. Thus, to date, the technology has not progressed from an era of narrow, learning-only AI that could pass the Turing test to a general AI machine that can operate autonomously, without prior knowledge, in a manner similar to human reasoning. And we are not even talking about super-intelligences that are on a par with humans, or even superior to them in terms of cognitive abilities. (Such an era would mean the achievement of the singularity, in which machines and humans would appear as equals in the world). But to sum up, one of the most important factors of our time is artificial intelligence. AI tools are usually at the heart of innovation, and this is no different in the legal sector, where more and more LegalTech tools are based on AI systems. Nevertheless, there are also many risks inherent in this new technology, which make the lack of legal regulation in this area a critical concern. The EU AI Act, which has been adopted but has not yet entered into force, is therefore a major step forward.

The road to AI Act

One of the main objectives of the European Commission (as of April 2024), led by Ursula von der Leyen (hereafter: the Commission), is to create a greener, more digital and more equitable Europe. EU legislation on the digital space has already started with the Digital Services Act (DSA) Regulation and the Digital Market Act (DMA), the Data Act, which entered into force on January 11, 2024 , as well as the Data Governance Act. This includes the recently adopted AI Act, a comprehensive regulation on artificial intelligence systems. The EU has also taken steps on AI before the adoption of the AI Regulation. The Commission has set up a High-Level Expert Group to provide advice on the safe use of AI, and a White Paper on AI has been published. In 2021, the Commission presented a legislative proposal. The explanatory memorandum and the impact assessments carried out show that the main objectives of AI legislation are to ensure the safe use and transparent operation of AI systems from the user’s perspective and to avoid legal uncertainty for developers, replacing the soft-law instruments with actual legislative acts. According to the Commission, the justification for complying with the subsidiarity principle is that only common regulation in the regulated area will ensure sufficient efficiency and that Europe can become a world leader in the booming AI market. However, 2022 brought an unexpected turn of events for EU legislators, and the emergence of ChatGPT and other generative AI systems required a complete rethink of the previously considered draft regulations. At this point, regulation became ex post, post-technique legislation, which is also evidenced by the fact that the newly released AI systems are finally not included within the framework of the planned regulation, but in a separate chapter (Chapter V) under the title of general-purpose AI systems. Finally, on March 13, 2024, the European Parliament (hereafter: the Parliament) adopted the final text of the AI Regulation, which has not yet been published in the Official Journal but will enter into force 20 days later if it does, but will not become effective, as a rule, until two years after that date.

Legislative issues

The legislator already faced issues with the flexibility of the definitions and their corresponding precision at the time of drafting. A previous draft defined the scope of the Regulation, i.e. the definition of an AI system, by classifying systems using techniques defined in an annex as an AI system. This method is primarily conducive to a precise and comprehensible definition, but in the legislation adopted we are now faced with a rather complex formulation which, in my opinion, promotes a flexible legislative scope open to emerging techniques, but at the expense of the requirement for clarity of the rules, which is a requirement of legal certainty. This legislative challenge can be seen throughout the AI Act, as the legislator has not clearly stated which requirement it prefers, flexibility or clarity. In my view, in such a rapidly evolving area, it is more important to have flexible, open legislation, as a restriction to specific techniques may exclude some future techniques from the scope of the regulation, thus making it more difficult to achieve market uniformity in the area, which was one of the explicit aims of the legislation.  Nor, of course, does it seek a balance with the most important democratic principles of the rule of law, which also apply to the EU, including legal certainty.

In my upcoming piece, I will introduce the regulatory method that AI Act follows, and how does that affect the LegalTech tools that use AI.

Bertalan Máté Nagy – Law student at the Faculty of Law at the Eötvös Lóránd University in Budapest, Talent Scholar at the Aurum Foundation. In December 2023 acquired a certificate at an IT academy for lawyers, gained experience of coding, IT law, LegalTech. Member of CapsLaw digital law workgroup of the university, deals with IT law and LegalTech problems.