G7 to discuss AI cyber defence standards for financial institutions
Japan’s Satsuki Katayama warns banks must prioritise system fixes as advanced AI exposes cyber risks and Chinese models move closer.
- The G7 is preparing to discuss common standards for AI security and cyber defence tailored for financial institutions.
- The talks are prompted by the risk that advanced AI could expose a vast number of system vulnerabilities at once, overwhelming the capacity for immediate repair.
- A key focus will be on establishing criteria to help financial firms prioritize fixing the most critical vulnerabilities, such as those affecting deposit outflows or money laundering.
- The discussions will also cover closer cooperation among G7 nations to defend against cyberattacks, particularly those exploiting advanced AI models from countries like China.
The G7 is preparing to examine common standards for AI security and defence as financial institutions face the risk that frontier models could expose large numbers of weaknesses in their systems, Japanese financial services minister Satsuki Katayama told Jiji Press.
Katayama said banks and other financial institutions could not assume they would be able to repair every flaw at once if advanced AI tools identify vulnerabilities on a large scale.
They would therefore have to “decide the order of priority” when fixing their systems, she said.
The talks among the G7 major powers, including Japan, are expected to cover related criteria and closer cooperation against cyberattacks.
The concern has been sharpened by advanced models such as Claude Mythos, developed by US startup Anthropic, which are regarded as highly capable of finding system vulnerabilities.
Katayama has been negotiating with the United States to secure access to such technologies for major financial institutions in Japan.
She indicated that financial firms would need to identify the most critical areas in advance, rather than trying to tackle every vulnerability at the same time.
Priority areas could include measures to prevent deposit outflows and money laundering.
AI competition is currently led by three US companies: Anthropic, Alphabet Inc., the parent company of Google, and OpenAI, the developer of the ChatGPT generative AI chatbot.
Katayama acknowledged the existing gap, saying the “AI hierarchy” led by the United States had “already been established”.
However, experts believe Chinese companies are likely to close the gap over time.
Katayama stressed the need for Japan, the United States and Europe to cooperate in case advanced AI models released by Chinese companies are exploited and used in cyberattacks.
She also aims to act as a bridge between the United States and Europe at meetings of G7 finance ministers and central bank chiefs.
Japan’s Financial Services Agency is meanwhile considering using advanced AI models in cyberattack response drills at financial institutions.
Katayama said the agency would consult the National Cybersecurity Office, but added that the financial industry could hold such drills “ahead of other industries”.
[Copyright The Jiji Press, Ltd.]
