The head of the IMF cautions about advanced AI systems such as Mythos.

      TL;DR: IMF’s Georgieva cautioned that advanced AI like Mythos could "destroy the financial system" if misused and highlighted the risk of an AI bubble. She emphasized that addressing vulnerabilities incurs costs and that nations need fiscal capacity.

      The Managing Director of the International Monetary Fund, Kristalina Georgieva, warned on Wednesday that sophisticated AI systems such as Anthropic’s Mythos could potentially “destroy the financial system” if exploited incorrectly. These remarks were made in Brussels during the Fund's annual economic assessment of the eurozone.

      Georgieva indicated that “Mythos is only the beginning” and more similar models will emerge. She noted that while these advanced tools can swiftly identify cybersecurity threats “with speed and scale that was unthinkable until now,” they can also be misused. “In the wrong hands, that same ability could lead to the destruction of the financial system,” she stated.

      She urged countries to establish “all cybersecurity components” and called for enhanced collaboration between wealthy and developing nations, as well as between public and private sectors. Georgieva pointed out that there is no global cybersecurity organization and suggested that it’s hard to envision such an entity given the current geopolitical climate.

      She emphasized the need for advanced economies to assist the developing world in reinforcing its defenses, as the global financial network is too interconnected to ignore vulnerabilities. "If a significant weakness exists, it will be exploited," she warned. Furthermore, she encouraged nations to include AI-related cybersecurity expenditures in their budgets, stating, “Recognize that patching costs money. Ensure you have fiscal space.”

      Her warning coincides with central banks and regulators hastily responding to the risks posed by Mythos. The Federal Reserve and Treasury have met with bank CEOs, the ECB is conducting cybersecurity drills, and euro-area finance ministers have requested access to Mythos after realizing none of the EU institutions can examine its findings. Countries such as Japan, Australia, and the UK have also sought information on the matter.

      Georgieva also brought attention to the potential for the AI investment surge to shift into a downturn. “We might see the current AI excitement, the AI boom, at risk of becoming an AI bust,” she noted, adding that while the likelihood is low, it’s not nonexistent. Thus, it falls into the category of low probability but high impact risks.

      This caution arrives on the same day SpaceX debuted with a $2 trillion valuation, Amazon secured $17.5 billion for AI infrastructure, and Alphabet’s $85 billion equity offering concluded. Big Tech’s total AI-related capital commitment by 2026 exceeds $700 billion. Georgieva, with diplomatic intent, conveyed that the financial system is ill-prepared for scenarios where such investments do not yield returns.

      The eurozone's economic outlook has also weakened, with the IMF forecasting a 0.9% growth in 2026, which is down 0.5 percentage points from pre-war projections. Inflation is expected to rise to 2.8%. Georgieva mentioned that 80% of governments' initiatives to help households manage energy costs do not meet the IMF's criteria of being “temporary and targeted.” Given this context, persuading nations to allocate fiscal resources for AI cybersecurity might be challenging, yet it is a message the IMF is advocating nonetheless.