The European Central Bank (ECB) has published a Financial Stability Review analysis identifying dependence on a small group of artificial intelligence (AI) providers as an immediate operational risk for financial institutions. Wider use of similar investment models could amplify market shocks, while exposures to businesses undergoing AI transformation create additional channels for losses. The analysis calls for better data on AI use, stronger oversight of critical providers and adaptations to market safeguards and supervisory capabilities.

The share of euro area banks actively using generative AI rose from 46% in 2024 to 86% in 2025, although deployment remains less advanced in complex functions such as risk management. Reliance on a few US technology providers creates shared vulnerabilities, allowing outages, cyber incidents or model failures to affect multiple institutions simultaneously. In asset management, generative AI mainly supports research and operations, while established machine learning techniques play a more direct investment role. Equity funds using AI show a greater tendency than comparable funds to experience large losses simultaneously. These effects are unlikely to pose systemic risks at current adoption levels, but wider use of similar strategies could amplify stress through synchronised deleveraging and asset sales.

Financial institutions also face risks through their investments in firms whose earnings and competitive positions may change as AI spreads. Exposures are particularly significant through investment funds’ equity holdings, while bank lending remains more limited, suggesting that a reassessment would initially transmit mainly through market valuation losses rather than bank credit losses. Rising external borrowing by firms in the AI supply chain is also broadening potential losses beyond equity investors to creditors. Alongside stronger provider oversight, the analysis highlights the need to map shared dependencies, maintain exit strategies and strengthen governance as AI systems become more autonomous.