By Chuck Gallagher — Business Ethics Keynote Speaker and Trainer
TL;DR: Chuck Gallagher, AI ethics speaker and author, examines the ESRB’s July 2026 warning that frontier AI models now hand attackers a speed advantage in the EU financial system—and why the board, not the IT department, owns what happens next.
Picture a wire transfer. Thirty seconds, and it clears. The email that asked for it looked right, the signature looked right, and the urgency felt real enough that nobody stopped to ask a question. By the time someone did, the money was gone.
That scene isn’t science fiction anymore. It’s the risk European regulators just put in writing.
On July 7, 2026, the European Systemic Risk Board—the body that watches for cracks in the EU financial system—issued a formal warning. Its subject was frontier AI models and what they do to cyber defense. The board had already raised its read on systemic cyber risk from “elevated” in March to “severe” by June. That’s a short trip up a steep hill. I’ve watched this pattern take shape for a while now.
What Did the ESRB Actually Warn?
The board defined frontier AI models as advanced systems capable of materially shaping offensive or defensive cyber operations. Translation: the same tool that helps a defender can help an attacker. In the short to medium term, the ESRB says the edge goes to the attacker. These models let bad actors find weaknesses and run attacks with more speed, more scale, and more sophistication than before. Over the long haul, the same technology may strengthen defenses. But not yet.
The references the ECB itself points to describe AI as changing the economics of finding a vulnerability. What used to be slow and costly for an attacker is getting cheap and quick. The cheap thing for them becomes the expensive thing for you.
The ESRB flagged a second worry, too. Most of the companies building these frontier models aren’t in Europe. So the tools a European bank leans on are designed and updated somewhere else. The board calls that strategic dependency and geopolitical exposure. It means part of your defense runs on something you don’t control.
Why Is “Cyber Is IT’s Job” Now a Dangerous Idea?
Here’s the rationalization I hear in boardrooms. “Cyber? That’s IT’s department.” It’s a comfortable thing to believe. It’s also wrong, and the ECB just said so out loud.
The ECB sent its own letter the same day, straight to the CEOs of the significant euro area banks. The point was that these AI models have gotten fast enough to find a software vulnerability and write a working exploit for it in a fraction of the time it used to take, which leaves defenders with almost no room to react. Then the ECB said who owns the response. Primarily, it’s the bank’s management body—the board—and it doesn’t much matter at this point how long people have been calling this IT’s job.
“Cyber is IT’s job” becomes “cyber is everyone’s job” the first time an AI-accelerated scam empties an account before lunch. Trust me. The lesson always looks obvious after the loss.
What Does the ECB Now Expect From Bank Boards?
The letter came with dates attached. By October 31, 2026, each significant bank is expected to build a concrete action plan and hand it to its Joint Supervisory Team. The ECB named three priorities to move on first: faster vulnerability and patch management, sharper monitoring and AI-enabled defense, and a hard second look at third-party risk. It also pushed the annual IT Risk Questionnaire back from September 2026 to February 2027, so banks could put their attention where it matters now.
This lands on top of rules already in force. DORA—the EU’s Digital Operational Resilience Act—has required financial firms to report major technology incidents on a tight clock since early 2025. The first annual report came out in June 2026. It counted 3,383 major incidents across the EU financial sector in one year. Roughly a third came through third parties. About one in ten was a cyber event. Those are the numbers before you add a tool that writes exploits at machine speed.
DORA isn’t the only rulebook here. When a high-risk AI system is involved, the EU AI Act can require its own incident report, to a different authority, on a different clock. One filing doesn’t satisfy the other. A board can’t clear this with a single memo. The obligations stack up, and every one of them is running its own clock.
Why Do Boards Only Value Resilience After the Fire?
Now, the way I see it, this is an old human problem wearing new clothes. Boards treat resilience like insurance. They resent the premium right up until the house burns. Then they can’t sign the check fast enough.
I don’t say any of this from a high horse. I was an embezzler once. I stole from a client’s trust, lost my CPA license, and I paid for it in ways I am still paying for. Here is what stays with me: nobody was watching the account, so for a long stretch nobody knew. The boring, invisible work is the work that would have stopped me.
Frontier AI didn’t create this weakness. It just shortened the fuse. That’s the part I’d underline for any board reading the ESRB warning. The events don’t make the choice for you—you do. You just have less time to make it now. Every choice has a consequence.
Frequently Asked Questions
What did the ESRB warn about frontier AI models in July 2026?
On July 7, 2026, the ESRB warned that frontier AI models are changing how cyber threats reach the EU financial system. Near term, the edge goes to attackers—more speed, more scale, more sophistication. The board had already moved its systemic cyber risk read from “elevated” to “severe.” It wants national and EU authorities to build these risks into how they supervise.
Who is responsible for a bank’s AI-related cyber risk—IT or the board?
The board. In its July 2026 letter to euro area bank CEOs, the ECB put responsibility primarily with the bank’s management body—the board and senior management. The strategic calls on technology investment, resources, and risk tolerance are theirs, which means cyber isn’t something they can hand off wholesale to the IT department anymore.
What is the ECB’s deadline for banks to respond?
October 31, 2026. That’s when significant euro area banks are expected to hand a concrete action plan to their Joint Supervisory Team. The ECB also pushed the annual IT Risk Questionnaire deadline from September 2026 to February 2027, so banks could put resources on the priority areas first. Supervisors will review the plans and track progress from there.
How is AI changing cyberattacks—and can it help defenders too?
AI now lets attackers find software weaknesses and write working exploits faster than before, which compresses the time defenders have to react. As an AI ethics speaker and author, I’d point to the flip side the ESRB also noted: over time, the same technology is expected to strengthen defenses through better monitoring and detection. For now, though, the edge sits with the attacker.
Does this only affect European banks?
The rules are European. The exposure isn’t. The warning was written for the EU financial system, but the behavior it points to—treating cyber as a technical chore instead of something the board owns—shows up everywhere. As these attacks spread, any board that made that mistake is exposed the same way, wherever it sits.
Bring This Conversation to Your Organization
The organizations that come through this well won’t be the ones with the fanciest tools. They’ll be the ones whose leaders decided, early, that cyber resilience was a board’s job and a culture’s habit—not a line item somebody else owned. That’s a choice. And it’s the kind of choice I help boards and leadership teams make before the fire, not after. To bring this conversation to your organization, or to book a keynote on AI ethics and the consequences of the choices leaders make, visit ChuckGallagher.com.
Five Questions for Reflection
1.Where does “that’s IT’s job” still stand in for real board ownership here?
2.If an AI-accelerated attack drained an account before anyone noticed, what happens in our first ten minutes—and who actually moves?
3.Do we treat resilience spending as a premium we resent, or a habit we’ve already built?
4.How long, honestly, is the gap between a vulnerability showing up and our people responding to it?
5.By not deciding, what are we choosing right now
