In its latest report, the International Monetary Fund (IMF) has issued a warning to financial institutions that it must deploy sufficient counter measures to AI cyber attacks in a bid to avoid systemic escalation across other industries
The International Monetary Fund (IMF) has issued a warning to financial institutions that leveraging AI to automate systems leaves them vulnerable to cyber risks that could destabilise the entire industry.
The IMF highlighted the sophistication of AI in enabling cyber attackers to speed up their offensive capabilities, outpacing defences in the process.
AI cyber attacks could lead to heavy funding strains, solvency concerns and disrupts the overall market, the IMF said, particularly if financial institutions continue leveraging AI in today’s highly connective payments system.
In the IMF’s latest warning, it revealed several factors underpinning how AI tools can cause financial instability.
The first factor raised was that risks are systemic. As such, attacks can be exploited at scale rapidly until discovered. This is compounded further when the risk spreads across sectors, as many financial institutions have adopted similar cloud-based software to companies in the retail sector, for example.
The IMF reported it may only take one vulnerability within an AI system to cause a ripple effect to other institutions and industries. Reliances on a small number of software, cloud and AI models also pose risks and enhance the impact of weaknesses, claimed the IMF.
The culminating effect of these vulnerabilities could cause payment disruptions to consumers, while also affecting institutional liquidity.
“For financial authorities, the question is whether the system is prepared to absorb cyber incidents without destabilising core financial functions,” said the IMF in its report.
The Anthropic solution
To demonstrate the vulnerabilities of AI to cyber attacks, the IMF used the recent launch of Anthropic’s Claude Mythos Preview to showcase how generative AI companies are already working on detecting cyber threats in real-time.
The Mythos Preview is an AI model designed to assess the speed at which cyber threats or attacks are increasing in real-time, in a bid to help teams mitigate them before they can take effect.
Anthropic says its service can “find and exploit vulnerabilities in every major operating system and web browser”.
IMF believes models such as Mythos exemplify the challenge of risk management for AI-based cyber attacks, because they amplify existing techniques by operating at machine speed. Attackers have the advantage over defenders, argues IMF, because discovering and exploiting vulnerabilities can occur faster than patching and remediation.
In a financial system built on common software and shared service providers, this can create simultaneous vulnerabilities across many institutions.
Another solution outlined by the IMF was OpenAI’s GPT-5.5 version which assumes vulnerabilities will grow once detected, providing defensive tools to prepare for escalating threats whilst under appropriate governance.
AI’s “critical” counter-measure role
AI risk-based solutions are often the answer to combating AI cyber attacks, which the IMF believes is a critical part of the solution.
Financial institutions have been increasingly adopting AI tools to detect threats, prevent fraud, identify vulnerabilities, and respond to incidents, as outlined by IMF in its report.
AI can help reduce vulnerabilities at the development stage, rather than patching them after release. However, the IMF warns these benefits can only be truly effective if institutions invest in integration, governance, and human oversight; areas that relevant supervisors increasingly need to assess.
