IMF: AI tools like Mythos could spark global crisis

The International Monetary Fund warned in a blog post that AI models such as Anthropic’s Mythos could enable cyberattacks that interrupt payments, prompt insolvencies and strain liquidity, creating conditions that might trigger a global financial crisis.

The fund pointed to models that can identify software vulnerabilities at scale and to financial services’ reliance on shared cloud infrastructure. A single exploitable flaw could cascade across banks and other institutions and interrupt payment systems, the IMF wrote. Energy, telecommunications and public-sector systems use the same digital foundations, increasing the potential reach of a successful attack.

Anthropic released Mythos and created Project Glasswing to give the model to roughly 40 companies it judged critical to protect, including Nvidia, Apple, Amazon Web Services and Microsoft, rather than making it widely available. Anthropic’s CEO Dario Amodei stated the company plans to work with government officials to help defend the United States and its allies. British banks were given access to the model at the end of April, which prompted assessments by U.K. financial regulators.

The IMF noted that “cyber risk does not respect borders” and cautioned that smaller and emerging economies may be disproportionately exposed if international services are disrupted. The fund called for stronger international collaboration and regulatory measures to limit cross-border contagion from cyber incidents.

An independent analysis by Britain’s AI Security Institute found Mythos effective at detecting cyber vulnerabilities but warned that the greater concern is the proliferation of similar models. Researchers said future frontier models will likely be more capable, and recommended increased investment in defensive capabilities and oversight.

The IMF recommended that banks and other financial firms integrate AI tools into governance frameworks, maintain human oversight of automated systems, and strengthen business continuity and disaster recovery plans. It urged improvements in cyber and quality-assurance programs and routine cyber hygiene to reduce exposure.

Andy Ward, international senior vice-president at Absolute Security, warned that attackers are already leveraging AI to accelerate and scale threats. He added, “Without robust cyber resilience strategies and real-time visibility, the finance sector risks sleepwalking into deeper vulnerabilities.”

Regulators and banks are assessing the trade-offs of using large AI models to find vulnerabilities versus the risk those tools pose if misused. The IMF’s alert coincides with calls from central bankers and security officials for coordinated policy responses, stronger oversight and greater investment in cyber defense across public and private sectors.