Bundesbank Executive Board member Fritzi Köhler-Geib and economics professor Itay Goldstein recently discussed the opportunities and risks of AI-based trading as part of the Bundesbank Invited Speaker Series. The event was moderated by Falko Fecht, head of the Bundesbank’s Research Centre.

Risks posed by the autonomy of AI systems

In his introductory presentation, Itay Goldstein highlighted the transformative role of artificial intelligence (AI) in financial markets. AI agents using what is known as reinforcement learning are capable of learning from experience and dynamically adjusting their trading strategies. However, this autonomy carries risks: AI systems could develop unpredictable strategies that lead to market manipulation, price bubbles or even financial crises. One particular problem he mentioned is that interactions between AI agents often cannot be observed directly, making regulation more difficult.

Focus on collusion

Mr Goldstein presented his research findings showing that AI agents can develop collusive trading behaviours to maximise their profits. Collusion refers to behaviour whereby, for example, companies secretly work together to suppress or distort competition. AI agents could hold back information or manipulate prices, which would impair market liquidity and price discovery. Because they are capable of planning, AI agents could also deliberately create and exploit price bubbles.

According to Mr Goldstein, a comparison between AI and human traders shows that, while their computing power often gives AI systems the edge, human agents bring new information and intuition to the market. The interaction between these two groups is one of Mr Goldstein’s key research areas, as it has a major impact on market stability.

Panel discussion with Executive Board member Köhler-Geib

In the subsequent discussion with Bundesbank Executive Board member Fritzi Köhler-Geib and moderator Falko Fecht, participants dived deeper into the opportunities and risks of AI-based trading and the role of central banks. Ms Köhler-Geib stressed that although AI can make portfolios more efficient and diversified, it also poses risks. She underscored the importance of having a deep understanding of how AI systems function so that financial stability can be safeguarded.

Mr Goldstein suggested that regulators could use AI agents themselves to detect harmful behaviour. Ms Köhler-Geib noted that the Bundesbank is working together with institutions such as the Bank of England and the BIS Innovation Hub as part of the “Logos” project, which simulates the impact of AI on financial markets. She said that central banks should remain active and use modern technologies to fulfil their role as guardians of financial stability even in a world that is increasingly shaped by AI.

As part of the Bundesbank Invited Speaker Series, Bundesbank Executive Board members hold regular discussions with leading researchers and academics about current research and central bank-related topics. Most recently, Bundesbank Executive Board member Ms Köhler-Geib spoke with economics professor Darrel Duffie about the Fed’s payment system and balance sheet.