EBI is happy to announce a contribution of Alessio Azzutti (University of Hamburg and National University of Singapore), Wolf-Georg Ringe (University of Hamburg and University of Oxford) and H. Siegfried Stiehl (University of Hamburg) in the EBI Working Paper Series No. 130. Their paper entitled “The Regulation of AI Trading from an AI Life Cycle Perspective” was published on 28 October 2022.
Among innovative technologies, Artificial Intelligence (AI) is often avouched as the game changer in the provision of financial services. In this regard, the algorithmic trading domain is no exception. The impact of AI in the industry is a catalyst for transformation in the operations and the structure of capital markets. In effect, AI adds a further layer of system complexity, given its potential to alter the composition and behaviour of market actors, as well as the relationships among them. Despite the many expected benefits, the wide use of AI could also impose new and unprecedented risks to market participants and financial stability. Specifically, owing to the potential of AI trading to disrupt markets and cause harm, global financial regulators are faced today with the daunting task of how best to approach its regulation in order to foster innovation and competition without sacrificing market stability and integrity. While there are common challenges, each market player faces problems unique to the context-specific use of AI. In other words, there are no one-size-fits-all solutions for regulating AI in automated trading. Rather, any effective and future-proof AI-targeting regulation should be proportionate to the particular and additional risks arising from specific applications (e.g., due to the specific AI methods applied with their respective capability, validity and criticality). Therefore, financial regulators face a multi-faceted challenge. They must first define the additional risks posed by specific use cases that call for more in-depth scrutiny and, hence, identify the technical specificities that can facilitate the occurrence of those risks. Based on this assessment, they finally need to determine which AI characteristics require special regulatory treatment.
Inspired by the EU AI Act proposal, this paper examines the advantages of a ‘rule-based’ and ‘risk-oriented’ regulatory approach, combining both ex-ante and ex-post regulatory measures, that needs to be put in perspective with the ‘AI life cycle’. By advocating for a multi-stakeholder engagement in AI regulatory governance, it proposes a way forward to assist financial regulators and industry players – but even actors in public education – in understanding, identifying and mitigating the risks associated with automated trading through an engineering approach for the purpose of complexity mastering.