News 13/04/26

Human involvement and clear responsibility crucial when deploying AI in capital markets

AI is rapidly transforming the way trading in capital markets works. This creates opportunities: analyses are performed faster, costs decline and orders are executed more intelligently. But the same technology also reinforces existing vulnerabilities and can create new risks. In the report AI in Capital Markets: Balancing Innovation and Integrity, the AFM outlines a future in which AI models are able to operate so quickly and autonomously that they influence market behaviour in ways investors do not always realise. That is why we are initiating the conversation with the market and with national and international supervisors about opportunities and risks now.

In short

AI can influence price formation without prior agreement
Information becomes a new vulnerability
Rise of multi-agent technology
Looking ahead

AI can influence price formation without prior agreement

Self-learning trading systems can give rise to patterns that resemble coordinated behaviour, without parties having consulted each other or having intended to do so. These systems can mirror and reinforce each other’s price movements. Because many models use the same data and pursue the same objectives, the market becomes more vulnerable. An erroneous or defensive reaction by a single model can spread extremely rapidly when adequate controls are lacking, leading to price patterns that are no longer the result of healthy market functioning.

Information becomes a new vulnerability

An increasing number of trading models process not only prices, but also news, market sentiment, and information on social media. As a result, manipulation may shift from the market itself to the information on which those models run. A single, cleverly disseminated fake report could theoretically set hundreds of algorithms in motion within seconds, without a single misleading order needing to be placed.

Rise of multi-agent technology

At the same time, alongside autonomous AI systems, we are also seeing the emergence of multi-agent technology. This development offers opportunities for greater speed and efficiency, but above all requires careful design and appropriate risk management. Market participants must be agile enough to anticipate new applications. It is up to market participants to determine where technology is deployed, what boundaries apply and how they supervise risks and possible interventions. Through transparent choices, visible human oversight and clear responsibilities, confidence in the market remains safeguarded.

Looking ahead

The sector is demonstrating an active and aware attitude towards the increasing use of AI. In a rapidly changing technological environment, it is important that parties remain agile so new developments can be integrated in good time. At the same time, appropriate safeguards are essential to keep risks under control and remain in control.

We support this development by further strengthening our supervision and internal expertise. Risks are translated at an early stage into appropriate supervisory measures, with an eye for potential needs for additional interpretations or legislation and regulations. Recently, rules and expectations for algorithmic trading and the use of AI have also been clarified. Through continuous dialogue with the sector and in cooperation with national and international partners, we contribute to a stable and future proof market in which innovation and control are in balance.