Frontier AI Is Testing Operational Resilience in European Banking

Major banks directly supervised by the European Central Bank are now on the clock. By October 31, 2026, they must submit action plans showing how they will respond to the rise in AI-enabled cyberthreats. The plans must address faster vulnerability and patch management, stronger monitoring and detection, third-party risk, and operational resilience. Banks must show they can adapt to a faster threat cycle without disrupting the critical services they protect.

The ECB’s July 7 letter explains the urgency. Since DORA took effect in January 2025, frontier AI has intensified a familiar risk. Now, software vulnerabilities can be found and exploited at unprecedented speed, leaving banks with less time to identify, test, and deploy a fix safely.

“Because of frontier AI, agents can now find vulnerabilities and identify security issues very quickly,” said Aishwarya Ramani, senior solutions marketing manager at Illumio. “But they can also exploit those vulnerabilities at machine speed.”

That speed creates a difficult balance. Banks must patch, but a rushed change to a core banking system, payment platform, cloud environment, or critical third-party connection can disrupt the services they're trying to protect. They need to reduce exposure quickly without causing an outage of their own.

The vulnerability window is shrinking

Traditional vulnerability management followed a clear sequence: A flaw became known. Security teams assessed the exposure, tested a fix, and deployed the patch. That process never guaranteed safety, but it often gave defenders time to act. Frontier AI can compress — and potentially eliminate — that window.

CERT-EU warned that “AI-powered tools are now discovering and exploiting software vulnerabilities at a pace that fundamentally breaks the traditional patch cycle.” It also cited evidence that exploitation can begin before a patch is available.

That puts bank security and IT teams under two competing pressures. They must assess more vulnerabilities and respond to faster attacks. At the same time, they must test patches, keep systems stable, and keep critical financial services running.

Regulators see this as a lasting shift. In June, the Five Eyes cyber agencies warned that the timeline for major changes in AI-enabled cyber capabilities “is not years, it’s months.” The European Systemic Risk Board similarly warned that frontier AI may increase the speed, scale, and sophistication of cyberattacks against the EU financial system.

Together, these warnings make vulnerability management an operational resilience issue. Banks must still patch quickly, but they can’t rely on patches alone. When a fix is unavailable or can’t be deployed at once, they must reduce exposure, protect critical services, and limit what an attacker can reach.

“When something can’t be patched immediately, you can still shrink what it can reach,” Ramani said. “If something is compromised, you contain it in place instead of watching it spread.”

DORA is the framework for proving resilience

DORA has applied to EU financial entities since January 2025. Its five pillars cover information and communication technology (ICT) risk management, incident reporting, resilience testing, third-party risk management, and information sharing.

But DORA doesn't assume perfect prevention. It requires firms to identify the systems and providers that support critical services. Firms must also prove those services can withstand disruption.

The ECB’s October 31 deadline makes that test more urgent. Banks must show how their DORA programs will adapt to a threat cycle that leaves less time to patch. They must focus first on the most exposed systems while keeping core banking, payment, and customer services stable.

Dependencies expand the blast radius

A bank’s attack surface extends beyond its own data center. It includes cloud environments, software components, managed services, payment connections, and third-party providers. A vulnerability in any one of them can create a path to critical services.

“You don’t know the extent of that dependency on the third party,” Ramani said. “Mapping those interdependencies, I think, is the biggest challenge.”

The recent cyber incident at Jack Henry makes the issue tangible. The U.S. financial technology provider serves more than 7,200 banks and credit unions. In August, it disclosed that a vishing attack had reached a limited part of its internal, nonproduction environment and affected the personally identifiable information of fewer than 10 clients.

Jack Henry said no client-facing systems, core platforms, or daily processing services were accessed or disrupted. Its controls detected the activity, isolated the affected systems, and contained the incident without an outage.

The event didn't cause a widespread financial disruption. But it shows why banks must understand their connections to major technology providers — and why they must contain the impacted systems if an attack targets one of those providers.

As Ramani puts it, “An assessment will tell you if a vendor was secure in March, but it says nothing about what happens when that vendor’s own environment is compromised in September, and your connection to them is still live.”

How banks can buy time

When banks can’t patch at once, they need controls that reduce exposure while they test and deploy a fix. Ramani described three steps: “Assess, prioritize, and detect.”

Banks can act on those steps by:

• Mapping current connections across applications, workloads, cloud environments, and third parties

• Prioritizing internet-facing assets, critical services, and the paths that create the greatest exposure

• Limiting unnecessary east-west communication so one compromised system can’t reach everything around it

• Using segmentation or microsegmentation to protect critical assets and reduce the blast radius

• Isolating compromised workloads or risky third-party connections before lateral movement spreads

• Testing response plans against high-speed exploitation, ransomware, cloud outages, and supply-chain disruption

Segmentation doesn't replace patching. It can compensate when a patch is unavailable, needs more testing, or poses too much operational risk to deploy. The ECB specifically lists segmentation and, where feasible, microsegmentation as part of defense in depth.

Resilience must keep pace with the threat

By October 31, banks must do more than acknowledge the frontier AI threat. They must show how they will manage it.

“Strong action plans will not assume banks can patch every vulnerability before attackers exploit it,” Ramani said. “They will show that banks know which systems are critical, understand how those systems connect, and can limit lateral movement when an attacker gets a foothold.”

When threats move at machine speed, resilience takes more than faster work. Banks must ensure that one vulnerability can’t cause a wider operational disruption.

Ready to learn more? Explore breach containment for financial services.

‍