The most powerful AI models are turning from technological risk to a threat to financial stability. The Chairman of the Financial Stability Board (Financial Stability Board, FSB) Andrew Bailey warned the finance ministers and heads of central banks of the G20 that the impact of advanced AI on cyber risks was the most immediate problem for the global financial system.
We are talking about models on the border of modern capabilities that get increasingly developed autonomy, are able to solve complex problems on their own and have skills suitable for cyber attacks. According to the FSB, such systems can dramatically change the speed, scale and cost of cyber operations. As a result, intruders will be able to look for weaknesses faster and carry out more attacks at a lower cost.
The financial sector is particularly vulnerable due to the close relationship between banks, exchanges, payment systems and technology suppliers. One serious incident could spread between countries through shared infrastructure, cloud platforms and other service providers. High market concentration increases risk, as many financial companies can depend on the same technology platform.
The FSB advises to prepare for scenarios in which the failure will simultaneously affect several organizations or general technological service. Banks and infrastructure companies will have to improve vulnerability management, respond faster to incidents, and work out recovery in advance. In the most severe cases, organizations should be able to restore systems from actually “bare iron” without relying on damaged software environments and data.
The rise of AI capabilities simultaneously accelerates the work of advocates and researchers. The British National Cyber Security Center warned about the approach of the “wave of updates”: AI helps to find the accumulated vulnerabilities faster than developers and administrators are used to eliminating them. Financial institutions will have to update the infrastructure more often, and too fast changes themselves can create operational problems if the testing and recovery processes do not have time to rebuild.
Fears are already beyond theory. Modern AI systems are increasingly doing well to find weaknesses in real-world software, and automation reduces the gap between detecting a problem and trying to operate it. Recent results show how quickly AI changes vulnerability search and forces advocates to review the usual corrections cycle.
The Financial Stability Board simultaneously sees AI as a potential protection tool. In June, the organization offered financial companies 12 recommendations for the safe adoption of the technology, including cyber risk management and dependence on third-party providers. The final version of the FSB document is scheduled to be released in October 2026. Bailey called on governments to develop international rules for the safe production and implementation of the most powerful models.
We are talking about models on the border of modern capabilities that get increasingly developed autonomy, are able to solve complex problems on their own and have skills suitable for cyber attacks. According to the FSB, such systems can dramatically change the speed, scale and cost of cyber operations. As a result, intruders will be able to look for weaknesses faster and carry out more attacks at a lower cost.
The financial sector is particularly vulnerable due to the close relationship between banks, exchanges, payment systems and technology suppliers. One serious incident could spread between countries through shared infrastructure, cloud platforms and other service providers. High market concentration increases risk, as many financial companies can depend on the same technology platform.
The FSB advises to prepare for scenarios in which the failure will simultaneously affect several organizations or general technological service. Banks and infrastructure companies will have to improve vulnerability management, respond faster to incidents, and work out recovery in advance. In the most severe cases, organizations should be able to restore systems from actually “bare iron” without relying on damaged software environments and data.
The rise of AI capabilities simultaneously accelerates the work of advocates and researchers. The British National Cyber Security Center warned about the approach of the “wave of updates”: AI helps to find the accumulated vulnerabilities faster than developers and administrators are used to eliminating them. Financial institutions will have to update the infrastructure more often, and too fast changes themselves can create operational problems if the testing and recovery processes do not have time to rebuild.
Fears are already beyond theory. Modern AI systems are increasingly doing well to find weaknesses in real-world software, and automation reduces the gap between detecting a problem and trying to operate it. Recent results show how quickly AI changes vulnerability search and forces advocates to review the usual corrections cycle.
The Financial Stability Board simultaneously sees AI as a potential protection tool. In June, the organization offered financial companies 12 recommendations for the safe adoption of the technology, including cyber risk management and dependence on third-party providers. The final version of the FSB document is scheduled to be released in October 2026. Bailey called on governments to develop international rules for the safe production and implementation of the most powerful models.
