G7 finance ministers agree to run AI cyberattack drills on banks

Finance ministers agree on AI attack drills

On Thursday, 15 October, Japan's finance minister Satsuki Katayama told reporters in Bangkok that G7 finance ministers had agreed to run simulation exercises on cyberattacks boosted by artificial intelligence. The decision came at a meeting of finance ministers and central bank governors from the G7's advanced economies, held on the margins of the IMF and World Bank annual meetings in the Thai capital. According to Katayama, deputy finance ministers first discussed the idea of simulating advanced AI threats, and ministers then approved it. A Japanese finance ministry official who attended the meeting described the plan as theoretical exercises. The agreement followed a warning from the International Monetary Fund earlier that week that an AI-assisted attack on a critical service of a financial institution could shake confidence in the financial system.

What the simulations would test

The scenarios will be built by the financial authorities of the G7 countries, which include Germany, France, Italy, the United Kingdom, the United States, Canada and Japan. The exercises are meant to test how finance ministries and central banks would react to cyberattacks that use AI. The official said the results would be used to judge whether existing protocols work and to revise the relevant handbooks. Depending on the scenarios chosen, private financial institutions could also take part alongside the ministries and central banks.

IMF warning and the debate on guardrails

On Tuesday, 13 October, the IMF pointed to threats to growth and financial stability, citing a possible turn in the AI investment cycle, heavy borrowing and a long-running conflict in the Middle East. Speaking on a panel that included Anthropic co-founder Jack Clark, IMF chief Kristalina Georgieva said AI was already strengthening the global economy but could deliver negative shocks if productivity gains arrived more slowly than hoped. She stressed the need for safeguards:

If there are no safeguards, then the risks may exceed the opportunities.

Georgieva compared AI development to Formula One race cars running without barriers to protect spectators. She argued that industry self-regulation was suitable because governments are poorly placed to oversee a fast-changing sector, but that it should involve society. The countries with the largest AI sectors, including the United States, China and India, along with countries in the AI supply chain, should be part of the process, she said. Clark agreed that standards were achievable and pointed to early policy frameworks in the European Union and voluntary commitments in the United States. He said his company regularly tests its systems and publishes details of those tests.

Central banks and supervisors respond

On a separate panel, Bundesbank President Joachim Nagel said AI had the potential to increase global imbalances. He pointed to the scale of spending needed for the technology:

That is clear when one considers the large investments required for AI, both in hardware and software development.

Georgieva also said that people everywhere would need access to digital infrastructure, electricity and adequate regulation before they could benefit from the technology. Banking supervisors around the world are stepping up efforts to limit the growing risks linked to AI. The Bank of England is working with commercial banks in the United Kingdom on broad options for recovering from cyberattacks, and the central bank said it can no longer assume its own defences are immune to AI-assisted attacks. Experts, including the heads of the Bank of England and the IMF, have stressed the need to protect financial institutions.