EU Artificial Intelligence Act; a Step towards the World’s First Legislative Framework in the AI Industry

Background

In recent years, there has been a notable surge in worldwide legislative scrutiny of Artificial Intelligence (AI) technologies. Governments and legislative bodies worldwide have recognized the need for comprehensive regulations tailored specifically to AI, with the aim of fostering responsible and ethical development, deployment, and utilization of AI systems. While numerous regulatory proposals were advancing, many encountered delays and still remain non-binding. However, noteworthy progress in implementing AI laws has been made primarily by the European Union.

Years of efforts culminate as the EU institutions and legislators make significant efforts in enacting the groundbreaking legislative framework for AI, ever since the European Commission first published a proposal for a regulation laying down harmonized rules on AI in April 2021, aiming to improve the functioning of the internal market in conformity with Union values, followed by the approval of the Council of the EU on the version of the proposed AI Act which was published on the 6th of December 2022. ?

On the 14th of June, 2023, the European Parliament has taken a decisive step towards shaping the future of AI within the EU, after approving its negotiating position on the proposed AI Act, with 499 votes in favor, 28 against, and 93 abstentions. The significant number of votes in favor underscores the growing recognition among European Parliament members of the need for robust and comprehensive regulations in the field of AI that addresses the challenges and potential risks associated with AI technologies.

Following the European Parliament’s vote, the adoption of its own stance on the AI Act set in motion deliberations among the three branches of the European Union, namely the European Commission, the Council, and the Parliament, in order to reconcile the different versions of the AI Act through the “trilogue” process, which is the three-way negotiations between the aforementioned branches.

Amendments proposed by the European Parliament

The newly adopted text by the European Parliament has integrated various amendments to the Commission’s proposal, encompassing an expanded list of intrusive and discriminatory applications of AI Systems. Notably, the text incorporates specific provisions pertaining to transparency and safety in relation to foundation models and generative AI systems.

Firstly, members of the European Parliament (MEPS) have introduced a definition of AI that aligns with the definition put forth by the Organization for Economic Co-operation and Development (OECD). The definition proposed of an AI system is as follows:

“a machine-based system that is designed to operate with varying levels of autonomy and that can, for explicit or implicit objectives, generate outputs such as predictions, recommendations, or decisions, that influence physical or virtual environments.”

This revision expands upon the Commission’s initial proposal, which originally defined an AI system as a software operating within predefined human objectives. The amendment now broadens the scope of AI to include virtual environments as part of the AI systems definition. ?

In addition, the European Parliament also implemented a list of prohibited AI practices, such as:

● Real-time remote biometric identification systems in publicly accessible spaces, excluding cases authorized by law enforcement for the investigation of serious crimes.

● Post-remote biometric identification systems, with the sole exception of law enforcement for the prosecution of serious crimes following judicial authorization.

● Biometric categorization systems utilizing sensitive characteristics such as gender, race, ethnicity, citizenship status, religion, or political orientation.

● Predictive policing systems relying on profiling, location data, or past criminal behavior.

● Emotion recognition systems employed in law enforcement, border management workplaces, and educational institutions.

● Unfocused collection of facial images from the Internet or closed-circuit television footage for the purpose of constructing facial recognition databases.

?

Furthermore, the European Parliament includes additional high-risk applications that poses “a significant risk to an individual’s health, safety, or fundamental rights”. These applications will be subject to compliance requirements, including pre-deployment conformity assessments, adhering to technical and auditing standards, and implementing monitoring protocols. Companies developing or deploying AI systems will therefore be required to thoroughly document and assess use cases to ascertain the suitable risk classification. Hence, in the case of generative foundation models, such as GPT, there are additional transparency requirements; providers must disclose when content has been generated by AI, implement measures to prevent the generation of illegal content, and publish summaries of copyrighted data used for training purposes.

?

MEPSs have also introduced regulations that impose certain obligations on providers of foundation models, which are rapidly developing aspects of AI. These obligations require providers to ensure protection of fundamental rights, health and safety, environmental concerns, democracy, and the rule of law.

?

Last but not least, under the AI Act, natural and legal persons or groups of natural persons will have the opportunity to lodge complaints against the providers or deployers of AI systems through a newly established EU AI Office, which will have a legal personality and will be managed by EU member states. Reports will be addressed to “a national supervisory authority”, responsible for overseeing the effective implementation and continued utilization of the AI Act within their respective jurisdictions throughout the EU.

?

Implementation

The EU is currently in the process of regulating artificial intelligence (AI) and aims to have the first operational trialogue by July 23, 2023. The official AI Act is anticipated to be passed by the end of 2023. However, even though there are still several phases remaining before the new law takes effect, it is crucial to assess whether member states and specifically companies are adequately prepared for the implementation of this act.

Several major European companies, such as Renault, Heineken, Airbus and Siemens, have collectively voiced their concerns and criticism of the EU's proposed AI regulations. They have expressed their opposition by signing an open letter addressed to the European Parliament, Commission, and EU member states. The signatories consider the AI Act ineffective and could have adverse effects on competition, in addition to jeopardizing Europe's competitiveness and technological sovereignty.

While the EU AI Act endeavors to ensure that AI technologies are beneficial and safe, it is essential to consider the potential impact on businesses, particularly in terms of stifling innovation in an era of continued technological advancement. Therefore, striking a balance between regulation and fostering innovation is essential.

The concerns raised by prominent European companies highlight the need for careful consideration and further deliberation in shaping the AI Act. It is crucial to address their apprehensions to avoid unintended consequences that may impede technological progress and hinder Europe's global competitiveness.

The article is written by: Rita Rahme , Associate.