Nearly nine in ten significant euro area banks are already using generative artificial intelligence, as regulators warn the technology could create new risks for markets, cyber security and the wider financial system.
Christine Lagarde, President of the European Central Bank (ECB) and Chair of the European Systemic Risk Board (ESRB), said the growth of AI is testing regulators’ ability to monitor risks that can build up across multiple institutions, in a speech in Frankfurt on Wednesday.
Generative AI is widely used across finance, with a recent survey also finding seven in ten EU securities market firms expect to increase their investment in AI.
Lagarde said most current uses involve limited autonomy, but “AI agents” — systems that can pursue a goal with limited human direction — are beginning to take on more discretion in areas such as trading strategies and identifying weaknesses in systems that trading depends on.
Trading, cyberattacks and geopolitics
The risk of many firms relying on similar AI models could cause them to react in the same way to shocks and reinforce price moves, the ESRB’s Advisory Scientific Committee has warned.
Use of more autonomous “agentic” AI remains limited, with 5% of asset managers saying in a survey this year that they gave AI autonomous or semi-autonomous authority over investment recommendations or trades, the ECB chief said.
Lagarde also cited research in which an AI model placed in a simulated trader role acted on inside information and then hid its reasoning from its manager, and separate simulated-market research where AI trading programmes learned to collude without communicating.
On cyber risks, Lagarde pointed out that tests show how quickly frontier AI models have advanced, with the latest models completing every step in a simulated 32-step cyberattack that late-2025 models could only partly complete.
The time between an initial cyber exploit and widespread automated exploitation could fall from weeks to hours as AI advances, the ESRB has warned.
Lagarde said Europe also faces risks from reliance on a small number of frontier AI models developed mainly in the United States and China, including the possibility that access could be restricted by export controls.
A US export-control directive in June covering two advanced models led their provider to suspend access, causing an abrupt cut-off for Europe, with access to a general-use model restored weeks later while access to another model remained limited to organisations vetted by the US administration, Lagarde stated.
She noted the ESRB has warned that cyber defences around critical financial systems need to be reviewed and updated as frontier AI changes the threats they face, adding that authorities should ensure firms plan timely responses and work together where an attack could spread across the sector.
