The European Central Bank has ordered euro-area banks to submit action plans for frontier AI cyber threats by end-October 2026. ECB supervisory board chair Claudia Buch wrote to bank CEOs citing the risk of AI models that can find software vulnerabilities and write exploits at unprecedented speed. Anthropic's Claude Mythos model, which reportedly discovered thousands of severe vulnerabilities in major OS and browser codebases, is a key driver of concern. The European Systemic Risk Board simultaneously elevated systemic cyber risk to 'severe' and flagged dependency on non-EU AI providers as a geopolitical vulnerability. No formal sanctions back the ECB order, but plans may be used to rank and pressure laggard banks. French startup Mistral is among firms racing to offer European alternatives to Mythos.
