The European Central Bank has warned that the AI-driven technology stock boom could end in a sharp market correction, threatening financial stability across the euro area. The warning comes as banks become increasingly reliant on the tech firms, models and infrastructure powering the AI economy.
For banks, this reliance is becoming hard to ignore. AI can boost speed and efficiency, but much of the underlying technology sits outside banking, controlled by a small cluster of providers.
Regulators are taking notice. The EU’s Digital Operational Resilience Act (DORA) sets requirements for ICT and third-party risk, while the UK’s critical-third-party regime gives regulators direct oversight of providers whose disruption could affect multiple financial institutions. Minutes from the Bank of England’s AI Consortium show UK financial leaders discussing provider concentration and its impact on market stability. The group found that concentration “arises from underlying characteristics of AI provision”, particularly at the model and computing levels, where alternatives can be scarce.
The systemic trap
The concern goes beyond a single provider suffering an outage. If multiple banks rely on the same models or cloud infrastructure, a single disruption could affect multiple institutions simultaneously.
This makes the AI supply chain a unique operational threat. As banks build models into core workflows, replacing a provider becomes far harder than swapping traditional software.
This vulnerability is already playing out. UK banks’ restricted access to Anthropic’s Mythos model prompted the government’s AI adviser to call it a wake-up call, illustrating how banks can lose access to important capabilities through decisions made elsewhere.
The ECB has also highlighted the cyber risks created by frontier AI. In July, ECB supervisory board chair Claudia Buch warned that advances in AI have “potentially profound implications for the confidentiality, integrity and resilience” of banks’ technology systems.
Banks must now map where their AI dependencies lie and plan for what happens if a provider changes its terms, cuts off access or suffers an extended outage.
Maintaining institutional control
The sheer pace of adoption makes this urgent. Research from the Cambridge Centre for Alternative Finance shows 81% of surveyed financial firms are adopting AI at some level, with 40% at an advanced stage. Yet only 14% see AI as transformational to their overall strategy or competitive edge.
This gap suggests firms are rolling out AI across individual departments before aligning those deployments with their wider strategy. In the rush, uncoordinated adoption creates shadow dependencies that can sit outside central risk oversight, especially when models, hardware and data services all come from different third parties.
True resilience requires more than a backup server. Banks must audit their tech suppliers, reduce vendor lock-in, maintain viable alternatives and stress-test what happens when a key model goes dark.
Banks cannot realistically turn their backs on AI. But they can decide how much control they retain over the systems powering it.
The priority now is clear: ensure that an AI provider going offline does not turn into a banking problem.
The European Central Bank has warned that the AI-driven technology stock boom could end in a sharp market correction, threatening financial stability across the euro area. The warning comes as banks become increasingly reliant on the tech firms, models and infrastructure powering the AI economy.
For banks, this reliance is becoming hard to ignore. AI can boost speed and efficiency, but much of the underlying technology sits outside banking, controlled by a small cluster of providers.
Regulators are taking notice. The EU’s Digital Operational Resilience Act (DORA) sets requirements for ICT and third-party risk, while the UK’s critical-third-party regime gives regulators direct oversight of providers whose disruption could affect multiple financial institutions. Minutes from the Bank of England’s AI Consortium show UK financial leaders discussing provider concentration and its impact on market stability. The group found that concentration “arises from underlying characteristics of AI provision”, particularly at the model and computing levels, where alternatives can be scarce.
