Chaen.ai Work with me

Insight

AI Cyberattacks and the Risk to Trust in Banks

🗓 ⏱ 6 min read

中文日本語

AI Cyberattacks and the Risk to Trust in Banks

AI cyberattacks raise questions beyond direct losses. I examine bank trust, rapid withdrawals, and responsible communication for Japan.

My concern about AI cyberattacks extends beyond compromised accounts to the possibility of damaged trust in banks. Here, I separate the figures and examples in my post from the risk scenario I drew from them, then explain how I would use that distinction in corporate communication in Japan.

AI cyberattacks raise a question about confidence

In my post, I expressed concern that advances in AI-enabled attacks could contribute to a major economic crisis. That was a warning about a possible chain of events, not a statement that such a crisis had begun or was inevitable.

The scenario moves from an attack on a bank to anxiety among its customers, then potentially to withdrawals at other banks. What concerns me is the possibility that the consequences would extend beyond the institution initially affected. Looking only at the direct damage could leave that wider question unexplored.

However, each link needs to remain a separate proposition. An attack does not automatically establish that deposits have been stolen. Concern about one bank does not establish that customers at every other bank will withdraw their money. Treating the entire sequence as settled would turn a scenario into an unsupported conclusion.

I referred to the Lehman crisis to express concern about distrust spreading and money ceasing to circulate. I was not establishing that the causes, scale, or outcome would be identical. When I use a comparison that strong, I need to make clear what it is intended to illustrate—and what it does not prove.

Securities account takeovers are not bank deposit theft

My post cited approximately ¥740 billion in unauthorized transactions associated with securities account takeovers in Japan in 2025. That figure was a starting point for discussing threats involving financial accounts. It was not evidence that the same amount had been stolen from bank deposits.

The distinction matters when translating the argument for readers outside Japan. The figure concerns securities accounts, not a confirmed attack on a major Japanese bank. The amount of unauthorized trading also should not be casually relabeled as a different measure of loss.

My broader concern was that the automation of attacks should prompt consideration of threats to banks as well. But concern about a future target and an account of an existing problem are different types of statement. The securities account figure alone cannot establish that a major bank will inevitably be breached.

If I were sharing this argument internally, I would keep the following points attached to the number:

  • It refers to unauthorized transactions involving securities accounts.
  • It is not a figure for stolen bank deposits.
  • The extension to banks is part of the risk scenario.
  • The size of a number does not, by itself, prove a future outcome.

This is not an attempt to minimize the concern. It is how I would keep the discussion focused on the right questions rather than on an increasingly dramatic interpretation of a single figure.

Cash, deposits, and the limits of the comparison

I also contrasted approximately ¥117 trillion in Japanese banknotes with more than ¥1,000 trillion in deposits. My purpose was to consider a scenario in which depositors all wanted cash at once—not to declare that the difference itself demonstrated an immediate banking crisis.

Those figures do not establish whether a particular bank would be unable to make a payment. Nor do they justify a blanket conclusion about the condition of Japanese banks. The useful question for my argument is what happens if similar withdrawal decisions become concentrated within a short period.

In the post, I described account balances in terms of people trusting the numbers they see. That framing led me to consider not only whether a technical incident occurs, but also what customers believe the incident means for them.

Imagine receiving information about a problem at one bank. The crucial distinction in my scenario is whether someone treats it as information about that institution or as proof that their own bank is experiencing the same problem. I am not claiming that everyone will make the latter judgment. I am identifying the point where the concern could spread.

For corporate communication, I would therefore prioritize the boundary of the statement: which institution, account, or service does the information actually concern? The comparison between cash and deposits is a reason to examine the scenario carefully, not a recommendation to withdraw money.

Digital withdrawals make speed part of the discussion

I mentioned Silicon Valley Bank because I was focused on the pace of withdrawals. In the post, I cited approximately ¥6 trillion withdrawn through smartphones in a single day in 2023.

I would not present that example as evidence of an AI-driven bank attack. Nor would I use the withdrawal figure alone as a complete explanation of the bank’s failure. Its role in this discussion is narrower: it illustrates the scale and speed of the withdrawals I was concerned about.

When considering AI cyberattacks, I would look beyond the time needed to investigate the technical problem. I would also ask what customers might hear, what remains unexplained, and what decisions they might make while that investigation continues.

This does not mean responding quickly by making claims that cannot be supported. It means preparing a clear structure for communicating uncertainty:

  • What has been established?
  • Which accounts or services does the statement cover?
  • What remains unknown?
  • What will be checked and communicated next?

I would rather define the limits of an update than imply that every question has already been answered. The concern about rapid withdrawals should make communication more precise, not make unsupported certainty acceptable.

Communicating AI risk to companies and audiences in Japan

For a Japanese corporate audience, I would not reduce this argument to a headline claiming that AI will destroy banks. The domestic securities account figure, the comparison between Japanese cash and deposits, and the overseas bank example serve different purposes. Combining them does not automatically turn them into proof of an approaching crisis.

I would apply the same standard to an overseas AI or technology company communicating with the Japanese market. If it uses a Japanese financial figure, it should retain the description of what that figure measures. If it uses an overseas example, it should not assume that Japan will experience the same outcome.

My preferred structure for a presentation or internal document would separate three kinds of content: the information cited in the post, the risk I infer from it, and the questions a company should examine. That lets the audience follow the reasoning without mistaking my interpretation for another reported event.

I also suggested in the post that more people might keep cash at home or place physical assets in safes. That was a possible behavioral response I imagined. It was not a report of a verified increase, and it was not a recommendation to move assets.

This distinction is particularly important to the purpose of the article. I want to examine how distrust could spread, not contribute to that spread by presenting an imagined response as something already happening. A warning becomes more useful when its scope is explicit.

Applying this to my own company

Within my own company, I would add a communication question to discussions of AI risk: how might users interpret the information they receive? Alongside the possible technical damage, I would examine how to distinguish confirmed facts from unresolved questions.

The practical lesson I take from this concern is not to predict the largest possible crisis. It is to preserve the meaning of figures, test each step in a risk scenario, and state clearly what an explanation does and does not cover. That is how I would connect a broad concern about trust to a usable corporate discussion.

FAQ

How could AI cyberattacks create a wider financial risk?

My concern is that fear about an attack on a bank could spread into doubts about other banks and trigger concentrated withdrawals. This is a risk scenario I raised, not a confirmed event or a prediction that a financial crisis is inevitable.

Does the ¥740 billion figure refer to stolen bank deposits?

No. In my post, that figure referred to unauthorized transactions associated with securities account takeovers in Japan in 2025. It should not be relabeled as the amount stolen from bank deposits.

Why did I mention Silicon Valley Bank?

I focused on the speed of withdrawals, citing approximately ¥6 trillion withdrawn through smartphones in a single day in 2023. I used it to discuss concentrated withdrawals, not as an example of a bank brought down by an AI cyberattack.

Based on a post on XOriginal post ↗

Share this article

XFacebookLINEはてブ

About the author

Masahiro Chaen
Masahiro Chaen Founder & CEO, DigiRise Inc.

CEO of DigiRise Inc. Has supported 600+ companies and 50,000+ people with AI training, consulting and development. GMO AI & Crypto advisor, Google Gemini AI advisor. 200K+ followers on X.