The Final Stage of the EU Legislative Process for the AI Act Has Begun

The European Parliament accepted its stance on AI regulation by an overwhelming majority on June 14, clearing the door for interinstitutional discussions to finalize the world’s first comprehensive law on artificial intelligence.

The AI Act is a flagship endeavour aimed at regulating this revolutionary technology based on its potential for damage. It takes a risk-based approach, prohibiting AI applications that represent an unacceptable danger and enforcing stringent rules for high-risk use cases.

The main focus of the last-minute attempts to amend the wording agreed at the parliamentary committee level was where to draw the line regarding the kinds of AI applications that should be prohibited. It is reported that the purpose of plenary changes is to send a political statement rather than to change the text.

Within the scope of the AI Act, the EU legislators established a tiered approach for AI models that do not have a specified goal, known as general goal AI, with tougher regulation for foundation models, and huge language models on which other AI systems can be constructed.

The first layer concerns generative AI, such as ChatGPT, for which the European Parliament intends to mandate the labelling of AI-generated material and require the disclosure of copyrighted training data.

Reporters were told that the Parliament wants to anticipate the legislation’s two-year application time, if not for all forms of AI, at least for foundation models or generative AI, considering the disruptive consequences these models are currently having.

The list of prohibited practices was extended to subliminal techniques, biometric categorisation, predictive policing, internet-scrapped facial recognition databases, and emotion recognition software is forbidden in law enforcement, border management, workplace and education.

An additional layer was introduced for AI applications to fall into the high-risk category, while the list of high-risk domains and use cases in law enforcement and migration control was refined and expanded. Popular social media recommender systems have been classified as high-risk.

Risk management, data governance, and technical documentation requirements for high-risk AI providers were strengthened. Assessments of the potential effects on fundamental rights and the environment were made subject to new regulations.

The Commission was given the important task of resolving conflicts between authorities, but an AI Office was set up to help collaborate on cross-border issues.

The members of the European Parliament will now engage in trialogue conversations with the EU Council of Ministers, which represents European governments, and the European Commission.

According to the interpretations, high-risk categories, basic rights, and foundation models are expected to be the key topics of debate. On the other hand, technical solutions to problems like governance, innovation, and AI definition are probably in order.

It is considered to be essential for companies operating in the field of AI to initiate compliance processes already, within the scope of the fiction designed by the EU.

You may access the details of the vote at the European Parliament session on 14 June 2023, and the AI Act here.

Tagged with: Gökçe, Görkem Gökçe, Elif Aksöz, Technology & Telecoms,Yapay Zeka, AI, Artificial Intelligence