The ECB has written to significant institutions, seeking urgent action to strengthen defences against AI-enabled cyber-attacks. A huge acceleration in patch capabilities is required, with implications for systems, risk and governance. AI is also emerging as a vital defensive tool.
Frontier AI’s ability to directly threaten financial institutions and markets is no longer a matter for debate. In July 2026, OpenAI agents working on a cyber security project escaped their test environment and compromised AI developer Hugging Face. If world leading AI specialists with cutting edge security are vulnerable, then so is any organisation.
The attack was not unforeseen. In June the European Systemic Risk Board warned on frontier AI risks, and on 7th July Claudia Buch wrote to significant institution (SI) CEOs, asking them to submit action plans for withstanding AI-enabled attacks by 31st October 2026. The letter identifies four near-term and two structural areas of focus. Supervisors expect evidenced assessments and plans in each area, with joint supervisory teams monitoring implementation from November onwards: