AI in action · Regulation
ECB: AI Has Tilted Cyber Risk Toward the Attackers
The ECB says AI has left bank cyber defenders "structurally disadvantaged, at least in the near term". About a fifth of incidents at euro area banks start at third-party providers.
The European Central Bank has put a name to a worry its supervisors have been raising all year. In "Decoding the cyber-AI nexus", a special feature dated November 16, 2026 for its November Financial Stability Review, four ECB economists argue that agentic AI has lowered the barriers to entry for cyberattacks, so that attacks are now limited "more by the financial resources of the attackers and less by their human expertise."
The balance of power is the core of the argument. Defenders, the authors write, are "structurally disadvantaged, at least in the near term": they carry deployment costs, governance requirements and slower decisions, and are constrained by the price of frontier models, while cheaper models are spreading quickly. On simpler cyber benchmarks, success rates are high across many models, and a Chinese model matched the performance of Mythos, a leading frontier model, four months after its release. Close to half of newly disclosed vulnerabilities are classified as critical or highly severe.
The bank data show a split rather than a general rise. In the first half of 2026, 56% of euro area banks had no attack that resulted, or could have resulted, in a major ICT incident, up from 49%. At the other end, 11% of banks saw 11 or more such attacks, up from 7%. The system has not yet had a systemic cyber event, and most major incidents reported under the EU's DORA rules had limited impact.
The systemic channel is concentration. Euro area banks "rely heavily on concentrated third-party providers", about one-fifth of their cyber incidents originate at those providers, and a single provider incident "could simultaneously affect large parts of the euro area banking system." ECB Vice-President Boris Vujčić made the same point in a Frankfurt keynote the same day, saying AI is "magnifying the scale, speed and sophistication of cyberattacks."
The proposals are mostly about coordination: joint planning between banks, AI developers, ICT and cybersecurity firms and authorities; firm-level "assumed compromise" defences; frequent testing; mapping common exposures and single points of failure; and full use of the AI Act and DORA.
Sources: European Central Bank — Decoding the cyber-AI nexus, Financial Stability Review special feature (Nov 16, 2026); European Central Bank — Boris Vujčić, Risks and resilience in the euro area financial system, Euro Finance Week (Nov 16, 2026)