Marknaden på autopilot : Artikel 12 MAR i ljuset av artificiell intelligens

Sammanfattning

Digitalisation has rapidly transformed society, not least the financial markets. As technology has developed, it has become easier than ever for individual investors to participate in the market. In many cases, a few clicks and a couple of minutes is enough to open an investment account and start trading. However, this added simplicity has opened new doors for market manipulation to take place. Simultaneously, artificial intelligence (AI) has become an integral part of modern financial markets, both as a tool for generating and disseminating information and as an autonomous trading agent. The EU Market Abuse Regulation (MAR) is largely constructed around the assumption that humans make trading decisions on the market, leading to new questions regarding manipulation being presented when machines enter the markets as well. This thesis examines to what extent the elements of Article 12 MAR can be understood and applied when market manipulation occurs through AI-generated information and AI-driven trading. The analysis shows that the wording of Article 12.1 a)-c) MAR is sufficiently broad and technologically neutral enough to cover behaviours generated by AI systems on both transaction-based and information-based side. However, AI challenges several underlying assumptions of the regulation. The speed, opacity and autonomy of AI systems make it more difficult to assess when signals are "misleading", to distinguish commercially motivated trading from manipulative behaviours, and to prove price impacts on the market. In particular, the traditional view of the reasonable investor and the idea of a functioning market integrity is strained in an environment where a substantial part of price formations result from interactions between systems rather than between humans. The thesis concludes that Article 12 MAR remains a viable framework for addressing market manipulation in an AI-driven environment, but that its effectiveness is conditional on how its key concepts are interpreted in light of new AI developments. In particular, the analysis highlights the issues of the continued use of a human reasonable investor in a market increasingly shaped by interactions between systems, as well as the tension between technologically neutral wording and human assumptions which are embedded in the regulation. These findings suggest that the main challenge is not that AI falls outside the scope of Article 12 MAR, but that the provision must be applied in a market structure that it was never fully designed to address.