AI and Market Abuse Regulation

The current debate on artificial intelligence (‘AI’) and its broad legal implications is being actively discussed in several fields, including financial law. One of the most fruitful topics regarding, in particular, capital markets law, is the impact of AI on market abuse practices and supervision. In our book Artificial Intelligence and Market Abuse Legislation. A European Perspective, Edward Elgar, 2023, we discuss how technological and legal developments in AI affect the understanding, application, and enforcement of the EU Market Abuse Regulation (Regulation No. 596/2014 of 16 April 2014, hereinafter ‘MAR’).  The subject is explored by analysing the two main areas of the Regulation: (i) on the one hand, inside information and the disclosure regime; and (ii) on the other hand, market manipulation.

Regarding the former, the book investigates whether AI systems can play a prominent role in supporting the internal procedures carried out by in-scope issuers for the mandatory identification and prompt disclosure of inside information. We elaborate on this first dimension of the relationship between the MAR and AI. We observe that due to their ability to handle large amounts of data and their self-learning capabilities, AI systems may serve as a useful means to support the issuer and its management body in the process that leads to the identification of information that might legally qualify as inside information. Similarly, AI systems may also assist in the procedures for the disclosure of inside information.

The analysis leads to the acknowledgement that AI systems might indeed be useful, inter alia, for large and complex issuers subject to the MAR disclosure rules and may play a significant role in corporate groups comprising multiple entities and subsidiaries. In such cases, inside information is usually harder to detect, monitor and disclose in a timely manner, especially in connection with multi-stage events. Deploying AI to assist in such activities, therefore, can well facilitate compliance with MAR obligations. We conclude, therefore, that the current regime, possibly enhanced by soft law instruments, should clearly support the introduction and development of AI systems for such purposes.

From a broader perspective, we also find that technological progress poses challenges to the traditional approach to qualifying inside information, as defined in Article 7 MAR. The book discusses the role of media and disputes the idea that the dissemination of information should necessarily be centred on the issuer.

Relatedly, the book explores how the use of AI systems may have significant impact on the liability of the issuer and its directors in relation to possible breaches of the disclosure regime. More specifically, a malfunctioning of the algorithm, resulting in delays, omissions, or inappropriate disclosures could amount to non-compliance with the applicable rules and entail liability risks to the issuer and, ultimately, to its management body.  The subject pertains to a non-harmonised area of EU law and therefore requires careful analysis of national regimes.

To such an extent, we first discuss the limitations and implications of potentially full automation of the process leading to the identification and disclosure of inside information. We conclude that, considering the current structure of the MAR, and its application under Member States’ national laws, AI systems should not entirely replace human action and intervention, particularly in the crucial steps leading to disclosure. As an alternative solution, the book takes the view that market abuse legislation should unequivocally establish that issuers must: (i) adopt appropriate organisational measures to identify and handle inside information, and (ii) identify the persons that are responsible for disclosing inside information.

Considering the second point of our research—ie, the relationship between AI and market manipulation—the book analyses the state-of-the-art on the long-standing and articulated debate on high-frequency trading and algorithmic trading. The main issue in this area is linked to liability and enforcement, particularly considering the emergence of truly automated trading systems. We therefore explain and elaborate on some of the most interesting proposals currently presented in the academic debate, delving into their respective features, advantages, and drawbacks. Some criticisms of the current Markets in Financial Instruments Directive approach towards algorithmic trading and high-frequency trading aside, the chapter finds that the existing regulatory landscape works quite well in respect of such technologies. While we do not call for major changes in the current rules, we nonetheless argue that there is room for two relevant improvements, namely:

1. a significant increase in the duties and powers of trading venue operators regarding the monitoring, detecting, and reporting of suspicious activity; and
2. the introduction of a specific obligation for investment firms using algorithmic trading and high-frequency trading systems, to include appropriate measures in their software to ensure that the algorithms are adequately designed to monitor, predict, and thus anticipate situations that may result in market manipulation.

The book concludes that the approach most likely to be followed by the forthcoming EU Regulation on AI—ie, based on the differentiation of risk levels and risk factors in relation to AI technologies—could also be considered as a reference for the relationship between AI and market abuse laws. While we are aware that the speed of the current developments may soon render our analysis obsolete, our hope is that at least some of our findings can be useful as a guide to the exciting, albeit complex, future that lies ahead.

Filippo Annunziata is Fellow Academic Board Member at the European Banking Institute in Frankfurt, Associate Professor of Financial Markets Law at Bocconi University and Professor of International Financial Regulation at Cà Foscari University.