The EU Parliament Has Adopted the AI Act, but Now the Hard Part Begins
The EU has earned accolades for being the first to pass a law on Artificial Intelligence. Nevertheless, the implementation of this law, particularly on smaller yet high-risk AI systems, is expected to be an arduous undertaking.
By Zsuzsa Detrekoi
The world is on the verge of implementing the first comprehensive AI law. On June 14, 2023, the EU Parliament overwhelmingly passed the Artificial Intelligence Act with 499 votes in favor, 28 against, and 93 abstentions. The primary objective of the Act is to safeguard health, safety, fundamental rights, democracy, and the environment from the adverse effects of artificial intelligence systems. Simultaneously, it aims to support innovation and promote the uptake of human-centric and trustworthy artificial intelligence. This is the first-ever attempt globally to establish a comprehensive AI law. With the Parliament’s vote, the trilogue period begins, with the aim of reaching an agreement by the end of 2023.
The EU’s intention to regulate this area is evident from the fact that the Commission introduced the first version of the Act in the spring of 2021. Although artificial intelligence models have been around for a while, the explosion of Chat GPT six months ago highlighted the urgent need for a reliable framework for AI systems.
The proposed Act follows a risk-based approach and distinguishes between general-purpose AI systems, high-risk AI, and prohibited AI practices. General-purpose AI systems with limited risk must comply with transparency obligations that reveal to users that they are interacting with an AI system. Generative AI systems such as Chat GPT must disclose that the content was generated by AI, design the model to prevent it from producing illegal content, and publish summaries of copyrighted data used for training the model.
AI systems that negatively impact safety or fundamental rights are considered high-risk and must register and comply with specific requirements, including transparency, information to users, risk management, record-keeping, and provide human oversight while using the system. Technical documentation must also be drawn up before the system is placed on the market and kept up to date.
Prohibited AI systems pose an unacceptable risk and are considered a threat to people; therefore, they are banned. This category includes cognitive behavioral manipulation of people or specific vulnerable groups, such as voice-activated toys that encourage dangerous behavior in children, social scoring, which classifies people based on behavior, socio-economic status, or personal characteristics, and real-time and remote biometric identification systems, such as facial recognition.
While the EU aims to provide a single, future-proof definition of AI, the definition has changed radically by excluding software as the basis for the definition. Instead, AI is defined as a machine-based system designed to operate with varying levels of autonomy and that can generate outputs such as predictions, recommendations, or decisions that influence physical or virtual environments.
The European regulators have been lauded for their progressive approach towards AI, but the future of the technology remains uncertain. With AI systems being developed worldwide and accessible through the internet, the implementation of the European AI Act poses a significant challenge. The EU is still struggling to regulate big tech companies, and enforcing the Act on smaller high-risk AI systems seems like a daunting task. Blocking these systems may be a possible solution, but implementing it across 27 member states could prove to be a challenging feat.
Zsuzsa Detrekoi is a technology and law expert with the Media and Journalism Research Center.
For more data and research about trends in journalism, politics and tech, check Media and Journalism Research Center website.
