IMF: Advanced AI models could threaten financial stability

On May 7, 2026 the International Monetary Fund warned in a blog post that advanced AI models such as Anthropic’s Claude Mythos Preview and OpenAI’s GPT-5.5 could enable cyberattacks and make “zero-day” exploits commonplace, posing a threat to global financial stability and market functioning.

The IMF wrote that the newest models can bypass traditional financial guardrails and that offensive capabilities currently outpace defensive tools. The fund highlighted that Mythos, released in a tightly controlled setting, can identify and exploit weaknesses in major operating systems and web browsers even when its users lack specialized hacking skills.

The IMF framed the risk as systemic. Shared infrastructure — a small set of cloud platforms, widely used software and machine-learning toolchains that underpin payments and data flows — can create correlated failures that ripple across institutions and borders. The fund wrote: “Cyber risks are increasingly about correlated failures that could threaten financial intermediation, payments, and confidence at the systemic level.”

The blog did not quantify potential losses but cited prior studies that such episodes could escalate into solvency problems and wider market disruption. The blog noted developing economies are particularly exposed because they often lack the resources and defenses of wealthier countries.

As a policy response, the IMF recommended a “resilience-first policy framework” that treats cybersecurity as a core element of financial stability policy. IMF Managing Director Kristalina Georgieva reiterated the approach in an interview: “We are very keen to see more attention to the guardrails that are necessary to protect financial stability in a world of AI.”

U.S. officials signaled a domestic response. National Economic Council Director Kevin Hassett told an interviewer on May 6 that the administration has mobilized an all-of-government effort with private firms to stress-test advanced AI models before wide deployment. He described consideration of an executive order to create a clearer approval pathway for future AI systems and compared the proposed process to safety reviews for new drugs.

The IMF warned AI could increase concentration risk because a handful of platforms dominate cloud computing, software development and machine-learning toolchains. A single exploit affecting core services could affect many financial institutions at once and scale targeted intrusions into systemic incidents.

The fund noted defenses exist, including AI-driven threat detection and improved patching practices, but warned offensive uses of AI are maturing more quickly than defenses. It called for stepped-up international cooperation among regulators, financial firms and technology providers to set standards, share threat intelligence and ensure rapid recovery mechanisms.

The IMF added that risks extend beyond finance because financial services are tightly linked with other critical sectors. Successful AI-enabled attacks could spread to utilities and communications. The fund urged policymakers to prioritize resilience and coordinate cross-border responses to limit the chance that a sophisticated cyberattack becomes a global market crisis.