ECB Summons Major Banks Over AI Cyber Risks as Claude Found 1,000s of Flaws
Updated
Updated · Computerworld · May 26
ECB Summons Major Banks Over AI Cyber Risks as Claude Found 1,000s of Flaws
11 articles · Updated · Computerworld · May 26
An emergency ECB meeting warned major banks that advanced AI models can uncover and exploit software weaknesses far faster than traditional hacking methods.
Frank Elderson told lenders they must install security updates much faster, as supervisors see AI sharply compressing the time between a flaw being found and being abused.
Anthropic's Claude Mythos Preview illustrates the risk: the company says the model identified thousands of serious vulnerabilities in operating systems and web browsers.
European banks are seen as more exposed because many lack access to such tools, while some US banks are already testing them and the ECB hopes they will share lessons.
When AI can reverse-engineer security patches in minutes, how can any bank truly be considered secure?
An AI creator refuses to patch a core flaw. Who is liable when the next financial crisis hits?
Machine-Speed Vulnerabilities: How Anthropic’s Claude Mythos AI Forced the ECB and Global Banks Into Emergency Cybersecurity Overdrive
Overview
In May 2026, the European Central Bank (ECB) urgently responded to a surge of AI-driven cyber threats, known as the 'Mythos shock,' after Anthropic’s powerful Claude Mythos AI model exposed major vulnerabilities across the financial sector. The ECB summoned 111 of the largest eurozone banks, highlighting that while some major Wall Street banks had access to Mythos, most European banks did not—a gap deemed 'unfortunate' by Vice-Chair Frank Elderson. Despite this, Elderson warned that cybercriminals could soon obtain such advanced AI tools, urging all banks to act quickly and not delay preparations for this new threat landscape.