Artificial intelligence (AI) is a fast-evolving technology, which can be adopted in an array of fields, such as healthcare, manufacturing, transport and energy, bringing forward multiple economic and societal benefits across Europe. However, while the EU has recognised the centrality of AI in the digital transformation of society, it has also highlighted that “the same elements and techniques that power socio-economic benefits of AI can also bring about new risks or negative consequences for individuals or the society”.
In light of these considerations, the EU has underlined the need for strong regulation to safeguard AI providers and deployers by providing them with straightforward requirements and obligations, without unduly hindering technological innovation and competitiveness in this sector.
In December 2023, the Council of the European Union and the European Parliament reached a provisional agreement on the proposal for the regulation on harmonised rules on Artificial Intelligence, the so-called Artificial Intelligence Act, which was firstly presented by the European Commission in April 2021. The AI Act is the first legislative proposal of its kind in the world and is, consequently, destined to set a global standard for AI regulation.
Carme Artigas, Spanish Secretary of state for digitalisation and artificial intelligence, has defined this agreement “a historical achievement and a huge milestone […]”, which joins a wider package of policy measures to promote the development and the circulation of safe and trustworthy AI in conformity with EU existing law on fundamental rights. The regulation also aims at encouraging investment and innovation in AI.
Application and Implementation
The AI Act will be implemented in the twenty-seven Member States of the European Union. However, there are exceptions to its application, as the regulation clarifies that it will not interfere, in any case, with the Member States’ exclusive competence in national security. Systems used exclusively for military and defence purposes will therefore not be subjected to the new AI legislation. At the same time, the regulation will not apply to AI systems used for the sole purpose of research and innovation, or to people using AI for non-professional reasons.
Prohibitions and rules for high-risk systems
Among the main objectives of the AI Act, the need to ensure that AI systems are safe and do not cause serious fundamental rights violations was at the heart of the negotiations. To this end, the regulation introduces a classification system that estimates the level of risk an AI technology potentially poses to the health and safety of an individual.
Consequently, AI systems presenting only limited risk will be subject to very light transparency obligations, while a wide range of high-risk AI systems will be authorised, but subject to a set of requirements and obligations before they gain access to the EU market. However, certain high-risk AI applications will be banned from the EU altogether because they are deemed threatening to citizens’ rights.
“The provisional agreement bans, for example, cognitive behavioural manipulation, the untargeted scraping of facial images from the internet or CCTV footage, emotion recognition in the workplace and educational institutions, social scoring, biometric categorisation to infer sensitive data, such as sexual orientation or religious beliefs, and some cases of predictive policing for individuals”.
The Commission’s proposal was revised to include specificities for law enforcement authorities that deploy high-risk AI systems and tools to carry out sensitive operations and activities. However, an effective mechanism has been introduced to ensure the ever-present protection of fundamental rights against potential misuses of these high-risk systems.
Moreover, when it comes to the use of real-time remote biometric identification systems in publicly accessible spaces, the provisional agreement defines the objectives “where such use is strictly necessary for law enforcement purposes and for which law enforcement authorities should therefore be exceptionally allowed to use such systems”.
Rules on general-purpose AI models
The proposed legislation takes into account the so-called general purpose AI, that is general-purpose AI (GPAI) systems that can be used for varying purposes. Foundation models are capable of efficiently completing a wide range of performative tasks, such as “generating video, text, images, conversing in lateral language, computing, or generating computer code”. The provisional agreement states that foundation models must comply with specific transparency obligations before they are introduced on the market. For high-impact foundation models, a stricter regime was outlined since these models are more complex and advanced, deploying larger amounts of data and, therefore, increasing the probability of risks.
System of governance and penalties
In addition, the AI Act provides for the creation of a common European AI governance system, through the establishment of an AI Office within the Commission as part of the administrative structure of the Directorate-General for Communication Networks, Content and Technology. The European AI Office will be the centre of AI expertise across the EU, by promoting the development and use of trustworthy AI. Alongside the AI Office, an AI Board will also be set up as a coordination platform and an advisory body to the Commission. The main purpose of the Board will be to contribute to the implementation of the AI Act in the Member States.
In addition, the AI Act includes non-compliance penalties with fines ranging from 7.5 million to 35 million euros, which vary depending on the type of infringement and the firm’s size.
Next steps
On Wednesday 13th March 2024, the European Parliament approved the AI Act with 523 votes in favour, 46 against and 49 abstentions. The regulation is still subject to a final lawyer-linguist check but is expected to be adopted before the end of the legislature, in May. The law also needs to be formally endorsed by the Council.
The AI Act has generated some criticism since it introduces stricter rules and regulations that might hamper European competitiveness in the sector and, consequently, lead other countries, like China and the USA, to dominate the industry as these countries have not yet issued a comprehensive regulation on AI.
The challenge for EU legislators has been to deliver a balanced legislation whose aim is to regulate AI but without curbing innovation potential. Experts in the field have raised concerns and doubts relating to certain aspects of the new legislation. The EU will consider the constructive suggestions provided and improve the legislation accordingly.
Sabrina Bianchi
Sabrina Bianchi is a graduate student in International Relations at the University of Bologna. She is currently working in the R&D Department of Maggioli Group, specialising in dissemination, communication and exploitation of European projects in the fields of Immersive Technologies, Sustainable Energy and Artificial Intelligence.
