G7 plans simulations of cyberattacks carried out by artificial intelligence

G7 finance ministers and central bank governors agreed on October 15 in Bangkok to organise simulations of cyberattacks against the financial system carried out with cutting-edge artificial intelligence models, known as “frontier” models. This was reported by Journal du Coin, drawing on the communiqué published by the French Ministry of the Economy and Finance and on statements by Japan’s finance minister gathered by Reuters. The meeting was held on the sidelines of the IMF annual meetings, chaired by Roland Lescure and the governor of the Banque de France, Emmanuel Moulin.

What the communiqué says, and what it does not

The official text is brief on the subject. According to the communiqué from the Ministry of the Economy and Finance, G7 ministers and governors “discussed the impact of frontier AI models on cyber risks in the financial sector and ways to strengthen collective preparedness”. The communiqué itself therefore only refers to a discussion, with the rest of the text covering Ukraine and fertilisers.

The details of the exercises come from Tokyo. According to Satsuki Katayama, quoted by Reuters, the G7 ministers’ deputies had been working for weeks on “tabletop exercises” devoted to frontier AI, and the agreement “was endorsed at ministerial level”. These are tabletop simulations: an attack scenario is played out before decision-makers, who must respond as if the situation were real.

How these simulations will work

According to the source, the G7 finance ministries and central banks will write the scenarios themselves. The exercises are meant to test existing protocols and to rewrite response manuals. Private banks may be involved depending on the nature of the simulations. No timetable has been announced: the agreement sets out an intention and a method, but not yet any dates.

This initiative follows a warning from the IMF. The Fund had cautioned, the same week, that an AI-assisted cyberattack targeting a critical service of a financial institution could be enough to shake confidence. The Bank of England has taken a similar line: according to Journal du Coin, its deputy governor is working with banks on recovery options following an attack, on the assumption that current defences are no longer safe from an AI-driven offensive.

Stablecoins: a file left to the FSB and the G20

Crypto-assets do not feature in the G7 communiqué. The stablecoin issue is moving forward on a different track, that of the Financial Stability Board (FSB), the body created by the G20 in 2009 to coordinate financial regulators and chaired since 2025 by Andrew Bailey, governor of the Bank of England.

In a letter sent on October 8 to G20 finance ministers, Andrew Bailey makes stablecoins “a key focus” of the FSB’s year. A note on their cross-border vulnerabilities has accompanied it since October 12. According to FSB figures cited by the source, the market is worth around 300 billion dollars as of mid-2026, and the dollar backs 99% of that value. Andrew Bailey writes that cooperation between regulators “is still at an early stage” and that national frameworks “remain untested under conditions of severe stress or insolvency”.

The FSB announces three workstreams: indicators for monitoring cross-border flows, better information sharing between authorities, and work on managing the failure of an issuer. These workstreams do not yet have a date. The FSB also published, on October 15, the final version of its good practices for the adoption of AI in finance, and Andrew Bailey asked that the topic remain on the agenda in 2027.

What scope for financial centres outside the G7?

The source mentions neither Switzerland nor any other financial centres outside the G7. Nor does it say whether authorities or institutions outside the group will be involved in the exercises. Generally speaking, a crisis simulation organised by a group of authorities focuses first on their own protocols and their own institutions.

The mechanism described by the IMF, an attack on a critical service that shakes confidence in a major institution, is nevertheless not unique to G7 countries. For a financial centre outside the group, the main lesson lies less in the exercise itself than in the signal it sends: G7 authorities now regard attacks carried out with artificial intelligence as a risk to be prepared for collectively. It remains to be seen, once the scenarios are written, what will be made public.

Sources

• Journal du Coin: “G7, AI and stablecoins: the 3 stablecoin workstreams ignored by finance ministers”