Draft of EU AI Act – Positive and negative key solutions and their practical implications

Lately, a #DraftofEUAIAct (Act) has rightfully been in the center of attention and a lot has been said and discussed on this issue, due to the fact that this is the premier attempt to regulate quickly emerging #AI technologies applications in everyday business in a comprehensive manner, which is the trait of the Civil Law system, dominant in Europe.

In this article, we will point out the key positives, as well as the negatives in the proposed draft. We will also try to answer what does it mean for Serbia and companies from Serbia dealing with #AI.

Key positives features

One of the main positive features of the Act is transparency. #DraftoftheEUAlAct envisages that the providers of generative #AI technologies (such as ChatGPT) are obliged to comply with transparency obligations in the regulation (ensuring users are informed the content was machine-generated)

Another main positive feature is the establishment of prohibited #AI practices and foreseen restrictions on unethical and discriminatory uses of #AI systems such as:

“Real-time” remote biometric identification systems in publicly accessible spaces for the purpose of law enforcement;

 “Post” remote biometric identification systems, with the only exception of law enforcement for the prosecution of serious crimes and only after judicial authorization;

 Biometric categorization systems using sensitive characteristics (e.g. gender, race, ethnicity, citizenship status, religion, political orientation);

Predictive policing systems (based on profiling, location or past criminal behavior);

Emotion recognition systems in law enforcement, border management, workplace, and educational institutions; and

Indiscriminate scraping of biometric data from social media or CCTV footage to create facial recognition databases (violating human rights and right to privacy).

Obligations of providers of generative #AI technologies to apply “adequate safeguards” in relation to the content their systems generate; and provide a summary of any copyrighted materials used to train their AIs.

Key negative feature(s)

Vagueness of definitions and obligations which most likely stems from the political agreement which was necessary to even put together this draft.

As it is now perfectly clear – for the developers of foundation models and general purpose AI, the compromise amendments of the EU Parliament and their general vagueness result in extremely extensive documentation obligations, exactly due to the fact that relevant legal definitions and obligations are vague.

Foreseen legal frame with its broad and vague definitions provided in the draft of the EU AI Act, which are the result of the political agreement within the EU Parliament rather than the straight professional legal approach to the matter at hand, have thus resulted in significant obligations and extra vigilance imposed to the developers.

One other consequence of this general broadness and vagueness of definitions contained in #DraftoftheEUAIAct is that it may be difficult to establish when the rules have actually been violated.

All this means that the developers will have to put in extra effort to secure the evidence of their work in order to be able to show (later on) that it was executed in accordance with the best available practice and legal guidelines.

This will require expert legal advice on how to identify key issues, establish the legally acceptable „algorithm“ of the work process, and execute the said process properly all with the goal to secure relevant evidence which will later on serve as proof that the prescribed legal requirements have been dully fulfilled.

Serbian perspective

Considering that Serbia is an EU member candidate country, it is expected that this legal framework (in its final form and substance which will be adopted in the foreseeable future) will be followed in Serbia, as the EU harmonization process is ongoing.

Also, many Serbian companies from the tech sector closely cooperate with EU-based companies, which means that they will have to immediately (once the Act is fully adopted and implemented in the EU) start to follow this matter and implement the necessary steps in order for their work to be compatible and acceptable by their EU partner companies.

This means that, once the #AIAct is fully adopted and implemented in the EU, Serbian tech companies will have to comply with it and implement the same legal „algorithm“ of the work process and other measures. However, it is better not to wait until the last minute and to start in due time with preparations even now.