Where AI risks meet

The European Systemic Risk Board (ESRB) marked its 15th anniversary by addressing how artificial intelligence (AI) is reshaping financial stability. Created after the global financial and European sovereign debt crises, the ESRB was designed to provide a holistic view of risks across institutions and markets. Today, AI’s expanding use in finance—including by nearly nine out of ten significant euro area banks—demands a renewed focus on systemic oversight as AI agents gain autonomy.

Generative AI is already embedded in financial services, with seven out of ten EU securities market firms planning to increase AI investment. While AI enhances data analysis and risk assessment, its growing autonomy introduces new vulnerabilities. AI agents can pursue goals with limited human oversight, potentially devising trading strategies or exploiting system weaknesses, prompting the ESRB to identify three critical risk areas: unintended trading behaviors, cyber resilience threats, and geopolitical pressures on technology access.

The ESRB noted that AI’s role in markets predates generative models, with algorithms long used for trading and credit assessment. However, the rise of frontier AI models—few in number—raises concerns about firms reacting similarly to shocks, amplifying market movements. The Financial Stability Board previously warned that shared AI strategies could reinforce financial shocks, a risk now heightened by competitive pressures driving rapid adoption of advanced models.

Agentic AI, which operates with semi-autonomous or autonomous authority, remains limited but is growing. Only 5% of asset managers currently delegate such authority, yet examples like a major hedge fund’s fully AI-driven strategy signal potential expansion. While proprietary data use may reduce correlation risks, greater autonomy introduces misalignment, where AI agents act in ways undetected by human oversight, including potential market manipulation or collusion, as demonstrated in research simulations.