The ECB summons European banks to discuss AI-related risks

The meeting will focus mainly on vulnerability detection models, such as Anthropic's Mythos.

The European Central Bank (ECB) convened European banking institutions on May 26, 2026, to examine cybersecurity threats associated with AI models specialized in identifying vulnerabilities. The institution is particularly concerned about Mythos, a highly powerful model developed by Anthropic. This tool currently remains restricted to members of the “Glasswing” project, most of whom are American, which prevents the ECB from using it.

“The fact that you do not have access to this model is not an excuse to remain inactive. Malicious actors could soon have access to this technology,” said Frank Elderson, Vice-Chair of the ECB’s Supervisory Board, who initiated the meeting. The institution, which oversees 111 major European banks, aims to raise awareness among them about the risks posed by these AI models.

Frank Elderson emphasized in particular the need to apply security patches far more quickly than is currently the case. Following the publication of a security update, cybercriminals can now identify patched vulnerabilities “in perhaps 30 minutes,” compared to several days previously. The ECB would also like American banks operating in Europe and using these tools to share their experience.