The finance ministries of the G7 have agreed to rehearse for an AI attack on the financial system. Finance leaders of the group agreed to hold tabletop exercises to guard against AI cyber attacks, Japanese Finance Minister Satsuki Katayama said on Thursday, in remarks reported by Reuters from Bangkok on October 16.
The agreement was signed off at a meeting of G7 finance ministers and central bank governors held on the sidelines of the International Monetary Fund’s annual gathering in the Thai capital. It follows an IMF warning this week that confidence could be undermined by an AI-enabled cyber attack on a critical service of a financial institution.
The G7 Agreement: Simulations For AI-Driven Attacks On Finance
Katayama traced the decision to work done below ministerial level. “The G7 finance deputies had been discussing the idea of doing tabletop exercises on frontier AI, which was signed off at the ministerial level,” she told reporters.
A Japanese finance ministry official who attended the meeting described the mechanics. The plan will feature simulations developed by G7 finance authorities to test how they can respond to AI-driven cyber attacks. The results will test the effectiveness of existing protocols and be used to revise the manuals authorities rely on to tackle such incidents.
The exercises will be run by G7 finance ministries and central banks, and could include private financial institutions depending on the nature of the simulations, the official added. No dates for the first exercise were announced, and the scenarios themselves have not been published.
The IMF Warning And The Bank Of England’s Cyber-Recovery Work
The G7 move sits inside a wider regulatory shift that Reuters has been tracking. Banking regulators globally are stepping up efforts to contain growing AI risks, with some pointing to the potential for cyberattacks on banks — or rogue agents — to undermine global financial stability.
The Bank of England is among them. It is working with banks on deep cyber-recovery options, its deputy governor said on Thursday, adding that banks can no longer assume their defences are immune to AI-enabled cyberattacks. That is a notable change of posture: the question for a major central bank is no longer whether defences hold, but how an institution recovers after they do not.
The IMF’s warning supplies the systemic framing. A successful AI-enabled attack on one critical financial service, the Fund cautioned, could damage confidence well beyond the institution hit — which is precisely the kind of cross-border contagion a G7 exercise is designed to practise against. The pattern matches the one in our report on the first incident playbook for AI agents published by NIST, the NSA and CISA: assume the incident, rehearse the response.
Why It Matters
The central entity is the tabletop exercise itself — a rehearsal, not a rule. Its attributes are coordination and revision: seven finance ministries and their central banks testing shared protocols against the same simulated attack, then rewriting their incident manuals from what the rehearsal exposes. That connects directly to the neighbours that matter here — the private banks that may be brought into the simulations, and the regulators already building parallel machinery, from the obligations set out in our EU AI Act explainer to the thresholds debated in the draft US AI safety bill Anthropic has been circulating.
For the AI industry, the signal is that frontier AI is now treated by finance ministries as an operational threat to plan around, in the same category as a conventional cyber incident — not a hypothetical to debate. Exercises of this kind also tend to produce concrete expectations for the institutions involved: tested recovery procedures, named responsibilities, and manuals that assume AI-speed attacks rather than human-speed ones.
Background
Tabletop exercises are a standard tool of financial crisis management: the G7 and its member central banks have long rehearsed responses to bank failures, payment outages and state-sponsored cyber intrusions by walking officials through a scripted scenario and grading the response. What is new is the adversary in the script. As AI systems have become capable of finding and exploiting software vulnerabilities at machine speed — and as incidents involving autonomous agents acting outside their brief have multiplied this year — the finance track of the G7 has taken up frontier AI as a standing item, moving this month from discussion among deputies to a ministerial decision to practise.
Source: Reuters
0 Comments