The world’s richest economies are preparing for a new kind of financial stress test: simulated attacks on banks powered by artificial intelligence. G7 finance ministers and central bank governors agreed to hold joint tabletop exercises after warnings that AI could sharpen cyber threats to the financial system, a move that reflects how fast the risk calculus around banking has shifted. Japan’s finance minister announced the plan in Bangkok, where officials gathered on the sidelines of the IMF annual meetings, underscoring that cyber defense is now a top-tier policy issue, not just a technology concern.
The agreement comes after the IMF warned this week that confidence could be undermined by an AI-enabled cyberattack on a critical financial-institution service. That warning gave new urgency to a longstanding threat: banks and market infrastructure can absorb routine digital attacks, but a coordinated AI-assisted strike could test response plans, communications and trust at the same time. The Bank of England’s deputy governor put the risk bluntly, saying banks can no longer assume their defenses are immune to AI-enabled cyberattacks.
Tabletop drills are not a live-fire event, but they are a serious rehearsal. According to Reuters, the G7 plan calls for simulations developed by finance authorities to test responses to AI-driven cyberattacks on the financial sector. The exercises will bring together G7 finance ministries and central banks, and may also include private financial institutions depending on the scenario. The point is not to predict the exact form of the next attack. It is to see where official playbooks break, where communication slows and whether banks, regulators and central banks can coordinate fast enough when an incident moves from theory to crisis.
Japan’s finance minister, Satsuki Katayama, told reporters on Thursday that the ministers and central bank governors had agreed to the plan. Reuters reported that the agreement was reached at the G7 meeting in Bangkok. Katayama said, “G7 deputy finance ministers had discussed plans for advanced AI-related tabletop exercises, and this was approved at the ministerial level.” The News International also quoted her as saying, “G7 finance deputies had been discussing the idea of doing tabletop exercises on frontier AI.” The Reuters version is the cleaner, authoritative account of the announcement.
The wording matters because it shows this is not simply a vague commitment to “do something” about cyber risk. It is a formal decision to build scenarios, test reactions and then feed the results back into official manuals for handling such incidents. Reuters said the results will be used to revise those manuals. That suggests a practical policy outcome rather than a symbolic statement. In other words, the G7 is not waiting for a breach to reveal the gaps. It is trying to find them first.
The timing also fits a broader shift in how officials talk about AI risk. For years, cyber threats were mostly framed around phishing, ransomware and data theft. AI changes the scale and speed of those threats. It can help attackers write better malware, automate reconnaissance and potentially target financial institutions with more convincing social engineering. The G7 plan does not claim to solve those problems. It does show that top finance officials now see AI as a force multiplier inside cyber warfare, especially when the target is the financial system’s plumbing.
Banks sit at the center of trust-based finance. Customers must believe deposits are safe, payments will clear and trading systems will keep functioning. That means a successful cyberattack can do damage far beyond the direct technical hit. If the market believes a bank, payment rail or core service is compromised, even a short outage can create confusion, liquidity pressure and reputational fallout. The IMF’s warning that confidence could be undermined by an AI-enabled attack on a critical financial-institution service is important because confidence is the real asset under threat.
That is why the G7 is involving both finance ministries and central banks. Ministries can coordinate policy and cross-border response, while central banks sit closer to payment systems, market operations and lender-of-last-resort functions. Bringing in private financial institutions, depending on the scenario, adds another layer. Real incidents do not stay neatly inside government boundaries. They spill into banks, custodians, payment processors and market utilities, all of which need to know who speaks first, who confirms the facts and who decides when to escalate.
There is also a political layer. The G7 is trying to signal that major economies are not treating AI risk as abstract ethics talk. They are treating it as operational resilience. That matters because the financial system runs on speed and confidence, and cyber incidents punish hesitation. A coordinated tabletop drill can expose whether an institution’s written procedure still works when an attack unfolds in minutes rather than hours. It can also show whether regulators in different countries can make decisions at the same pace when the pressure is global.
There were no asset, index or company price moves reported with this policy event, so the immediate market reaction is muted on the tape. But that should not be mistaken for a lack of relevance. The absence of a market swing reflects the nature of the story: this is a supervisory and resilience issue, not a earnings or tariff headline. For investors, the more important question is whether this marks the start of a more aggressive regulatory phase around AI-driven cyber readiness in finance.
The answer may depend on what the simulations uncover. Reuters reported that no dated execution timeline was specified, and the next concrete step is the development of simulated attack scenarios by G7 finance authorities. That means the news flow is likely to build gradually rather than in one dramatic event. If the scenarios reveal major coordination gaps, expect more discussion about incident reporting, cross-border communication and minimum cyber standards for banks and key market infrastructure. If the drills go smoothly, officials may still use them to tighten manuals and formalize response paths.
For banks and their investors, the risk is not just the attack itself. It is the possibility that AI lowers the cost of launching attacks faster than the industry raises its defenses. The Bank of England deputy governor’s warning that banks can no longer assume their defenses are immune to AI-enabled cyberattacks captures that reality. What used to be a technology problem is now a balance-sheet and policy problem too, because a weak response can turn a contained breach into a trust event.
The G7’s next step is to build the scenarios. That is where the real work begins. Officials will need to decide what the attack looks like, how far it spreads, whether it hits a single bank or a broader service provider, and how quickly the response has to move. They will also need to test the messy parts: public statements, incident verification, coordination between ministries and central banks, and whether private institutions can share enough information without creating more panic.
The bigger message is already clear. The world’s top financial policymakers are no longer discussing AI cyber risk as a future problem. They are rehearsing it now. That does not mean a crisis is imminent. It does mean the G7 thinks the next banking shock could arrive through code, not credit.
