IMF chief warns advanced AI models like Mythos "can be used to destroy the financial system"

During a presentation of the IMF's annual economic assessment of the eurozone in Brussels on June 12, 2026, International Monetary Fund (IMF) Managing Director Kristalina Georgieva delivered a stark warning regarding the potential weaponization of advanced artificial intelligence models. Georgieva specifically cited Anthropic's Mythos model, cautioning that such powerful AI could be leveraged to "destroy the financial system" if it falls into malicious hands. Georgieva elaborated that the same sophisticated capabilities that enable frontier AI models to rapidly identify cybersecurity vulnerabilities, a process previously unthinkable in terms of speed and scale, can also be exploited for destructive purposes. She stressed that the emergence of models like Mythos signifies a new era where the potential for systemic disruption is significantly heightened. In response to these escalating threats, the IMF chief urged all nations to ensure they have robust cybersecurity frameworks in place. She called for enhanced cooperation, particularly between wealthier and poorer countries, and a stronger partnership between the public and private sectors. Georgieva acknowledged the absence of a global cybersecurity organization, noting the challenges of establishing one in the current geopolitical climate, but underscored the imperative for collective action. Furthermore, Georgieva highlighted the interconnectedness of the global financial system, asserting that a significant weakness in any part of it could be exploited, impacting the entire system. She advised countries to allocate sufficient budgetary resources for AI-related cybersecurity, recognizing that "patching costs money" and requires adequate fiscal space. Her comments reflect a growing concern among international financial bodies about the need to proactively address the dual-use nature of advanced AI technologies and their profound implications for global financial stability.