WASHINGTON: The International Monetary Fund (IMF) warned on Thursday of the risks to global financial stability posed by cyber-attacks using advanced artificial intelligence tools and called for greater international cooperation on the issue.
“According to IMF analysis, losses from extreme cyber incidents could trigger funding strains, raise solvency concerns and disrupt broader markets,” the financial institution warned in a new report.
The study’s authors highlighted the risks posed by the highly interconnected nature of the global financial system, where advanced AI models can “dramatically reduce” the time and cost of exploiting vulnerabilities.
The warning comes weeks after AI company Anthropic warned that its as-yet-unreleased “Mythos” model was highly adept at finding and exploiting such weaknesses.
This model has been particularly efficient in identifying vulnerabilities that were previously unknown to developers and users. For hackers, these so-called “zero-day” vulnerabilities can be particularly dangerous.
White House economic adviser Kevin Hassett said Wednesday. fox News that there is a “whole-of-government” and private sector effort to test the model and ensure it does not harm U.S. businesses or the government.
A day earlier, the U.S. government announced a policy shift that would give tech giants access to new AI models to evaluate before they are released. The IMF warned that emerging and developing countries “often have more severe resource constraints and may be disproportionately exposed to attackers targeting weaker defenses.” The authors say the risks are systemic and cross-sector, involve the threat of contagion, and dependence on a small number of platforms and cloud providers is likely to increase “the impact if a single weakness is exploited.” “Defenses will always be breached, so resilience must also be prioritized, especially to limit the scope of an incident and ensure rapid recovery,” the report said.
IMF chief Kristalina Georgieva warned last month that the global financial system is unprepared for the cybersecurity threat posed by AI.
Published at Dawn on May 8, 2026
