| dc.description.abstract | Artificial Intelligence (‘AI’) has the potential to drastically alter legal theory and practice and EU competition law is no exception to this statement. This thesis focuses on how AI can be used for the benefit of competition law, with special regard to its effective public enforcement. The core research question that the thesis seeks to answer is as follows: ‘To what extent can AI assist in a more effective public enforcement of EU competition law, what are the potential challenges/risks of this endeavour and how can they be mitigated?’
Initially, after introducing a working definition of effective enforcement, it is maintained that effective enforcement is necessary not only in light of certain characteristics of the digital economy, but also in order to tackle other forms of anti competitive conduct emerging in other sectors.
As AI comprises various notions and there is no universal definition for it, the thesis utilises the definition of AI in the AI Act as a point of reference. Subsequently, some essential AI concepts are explained in order to comprehend how AI systems can actually improve competition law enforcement. After having provided this context, the typical stages of an EU antitrust public enforcement procedure and the effects of AI in each of these stages are demonstrated, followed by the exploration of certain AI tools that several national competition authorities (‘NCAs’) have developed or are planning to develop, in order to provide concrete examples of AI usage.
The thesis then moves on to answer the main research question, by analysing the use cases, advantages, limitations and risks of AI usage in the public enforcement of EU competition law. Specifically, it is argued that AI can engage in complex analysis of structured or unstructured data, detect patterns and anomalies in such data as well as monitor compliance with and evaluate the effectiveness of remedies. Nonetheless, it is also acknowledged that these positive effects are not limitless, considering certain issues, such as the possibility of false positives and false negatives and hallucinations and the potential lack of access to reliable and usable data.
After having illustrated the benefits and limitations of AI usage in the public enforcement of EU competition law, certain risks that may arise are discussed. Particularly, it is found that AI usage is liable to be associated with biases, privacy concerns and lack of transparency, explainability and accountability.
Overall, the thesis indicates that AI can materially ameliorate the antitrust public enforcement process, yet it is not suitable to tackle every potential issue that might arise in a case, nor it is always guaranteed that it will produce the most accurate outcome. For this reason, it rounds off by proposing recommendations to minimise the risks and limitations, mainly revolving around multidisciplinary internal or external teams within the NCAs and the European Commission, security and privacy safeguards, informed management decisions and human supervision of AI systems. | en_US |
| dc.subject | Artificial Intelligence, EU competition law enforcement, effective enforcement, bid-rigging, collusion, computational antitrust, data access, automated decision-making, bias, privacy, transparency, accountability, explainability, AI Act, GDPR, Data Act | en_US |
