Christine Lagarde warns AI could compound risks across Europe’s financial system

Christine Lagarde warned in Frankfurt on 1 October that artificial intelligence could compound risks across Europe’s financial system as banks and investment firms adopt more capable tools. Opening the European Systemic Risk Board’s annual conference, the European Central Bank president identified automated trading, cyberattacks and dependence on foreign AI providers as threats that authorities need to monitor together.

Her warning concerned possible channels through which problems could spread, rather than a finding that AI has already caused financial instability. Lagarde said nearly nine in ten significant euro area banks use generative AI. She also cited a survey in which seven in ten EU securities market firms responding expected to increase AI investment between 2025 and 2027. The technology can help institutions analyse data and assess risk, she said, even as its wider use creates new questions for supervisors.

How AI agents could affect financial market trading

Most current uses of AI in finance have limited autonomy, according to Lagarde, but agents are beginning to take on more discretion. Such systems could devise trading strategies with limited human direction. A survey she cited found that only 5% of asset managers had given AI autonomous or semi-autonomous authority over investment recommendations or trades this year. She described that use as limited, while warning that greater autonomy could make human oversight harder to sustain.

Lagarde pointed to the risk that firms using similar models might respond to a shock with similar trades, reinforcing a move in prices. She also raised the possibility of misalignment: an agent pursuing a goal in a way its human overseers neither intended nor detected. As illustrations, she cited research in which a model playing a trader at a fictional firm used an inside tip and concealed its reason, and a separate simulated market in which AI trading programs learned to collude without communicating. Those were research scenarios, not reported incidents of misconduct in live financial markets.

Why faster cyber threats concern the ESRB

The ESRB had already issued a warning in July about frontier AI and cyber resilience. Its release said the board assessed systemic cyber risk as severe in June, up from elevated in March. The board expected advanced models eventually to strengthen defences, but said they could give attackers an advantage in speed, scale and sophistication in the short to medium term. Lagarde cited an ESRB warning that the interval between an initial exploit and widespread automated exploitation could fall from weeks to hours.

Shared technology is central to that concern: an attack affecting a service used by several firms could spread beyond one institution. Lagarde said firms need time to test fixes before deploying them in essential services, while attackers can exploit a weakness as soon as they find it. Her argument was that supervisors should consider the financial system as a whole when assessing these threats.

Europe’s access to frontier AI models

Lagarde’s third concern was Europe’s dependence on frontier models developed chiefly in the United States and China. She said a US export-control directive in June led a provider to suspend access to two advanced models, including for European users. Access to the general-use model returned weeks later, she said, while access to the model with fewer cyber safeguards remained restricted to organisations vetted by the US administration. She said the interruption caused no discernible disruption to the financial system.

The episode nevertheless illustrated her concern about future reliance on a small number of providers. Lagarde recalled the faulty software update in 2024 that grounded flights and disrupted banking services worldwide as a precedent for concentrated technology risk. Her concern about a future loss of AI access was a scenario, not an account of a disruption caused by the June directive.

What Lagarde wants authorities to do

Lagarde called for careful monitoring of how the three risks might interact. She urged Europe to develop its own AI capabilities and said cyber defences around critical financial systems should be reviewed and updated. Financial authorities should ensure firms plan timely responses and cooperate where an attack could spread across the sector, she said. These were recommendations in her speech, not announcements that the measures had been completed.

The ESRB said in July that relevant authorities should reflect frontier AI risks in supervision and oversight. It said it would revisit the developments in quarterly risk assessments and consider further action within its remit if needed. Lagarde also called for global cooperation on powerful models. Neither her speech nor the ESRB release established when the risks she described might materialise, or whether any specific policy change would follow Thursday’s address.