When Bank Breaches Hit Home, Customers Pay First

When a bank’s systems are breached, the people who feel it first are not the executives briefing reporters. They are the customers whose savings sit behind those walls, the tellers and support staff who must explain to worried account holders why their access has been suspended, and the employees suddenly tasked with locking doors that were open yesterday. That is the human reality behind a piece of breaking news reported by Madison Mills at Axios: U.S. banks are taking hits from AI agents, and the findings have alarmed banking executives who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond.

Sam Altman highlighted the report on October 3, and the Axios reporting landed, as he noted, a day after scenarios of what such attacks could look like. For the people who work inside those banks, the alarm is not abstract. Restricting access to systems means new hurdles in daily workflows, security reviews that slow down ordinary tasks, and the quiet anxiety of professionals who know that the tools attackers are using are improving faster than their defenses.

The banking story is hardly the only one of its kind. On October 11, there were attacks on refrigeration systems in France that may have been facilitated by agents. Japan is experiencing a growing, ongoing wave of cyberattacks. And for months there have been attacks from OpenAI and others on places like Hugging Face and the Australian government. Each of these incidents, whatever its setting, ultimately lands on people: workers at the targeted organizations, residents who depend on the services affected, and the security staff who spend their nights responding.

Yesterday, The Information reported on what a Latin American hacking group called Cybercartel is up to, a report Sam Altman described as ominous. What makes this disconcerting is not just that hackers are stepping up their malicious use of agents. It is that they do not need state-of-the-art frontier models to cause mayhem. That point matters for ordinary people, because it means the threat does not depend on some distant technological breakthrough. The tools that already exist are enough to disrupt the institutions people rely on every day.

Altman argues this should serve as a wakeup call for those who focus all their attention on regulating frontier models, often measured in terms of how much compute they use. Even if we were to stop the development of future frontier models, he writes, we would still face serious problems from the misuse of existing models. In other words, the harm is not hypothetical and it is not deferred. Current models are already causing serious problems across society, in domains ranging from education to politics, through deepfakes and wholesale disinformation, to banking. Ignoring all of that in favor of speculative futures, he argues, would be a mistake.

The warning did not come out of nowhere. At the beginning of 2025, Altman warned in Politico that the world was about to face a perfect storm in cybersecurity, for four reasons. First, generative AI, rising in popularity and declining in price, is a perfect tool for cyberattackers. Although it is unreliable and prone to hallucinations, it is terrific at making plausible sounding text, such as phishing attacks designed to trick people into revealing credentials, and deepfaked videos, at virtually zero cost, allowing attackers to broaden their attacks. Already, a cybercrew bilked a Hong Kong bank out of 25 million dollars. Behind that figure are the people whose money or livelihoods passed through that institution.

Second, large language models are notoriously susceptible to jailbreaking and things like prompt-injection attacks, for which no known solution exists. Third, generative AI tools are increasingly being used to create code, and in some cases the coders using them do not fully understand the code written. That autogenerated code has already been shown, in some cases, to introduce new security holes, which means vulnerabilities can now enter systems through the everyday work of developers who may not realize what has slipped past them.

Finally, in the midst of all this, the new U.S. administration seems determined to deregulate as much as possible, slashing costs and even publicly shaming employees. This is where the human stakes become especially concrete. Federal employees who do their jobs may be frightened, and many will be tempted to look elsewhere. Enforcement and investigations will almost certainly decline in both quality and quantity, leaving the world quite vulnerable to ever more audacious attacks. For the civil servants involved, the choice between staying in a demoralized workforce and leaving a post they may care about is a personal one, with consequences for everyone who depends on their oversight.

As Altman puts it: well, here we are. The storm he forecast at the start of the year has arrived, and its effects are visible in bank security reviews, in attacks on infrastructure in France, in the wave hitting Japan, and in the activities of groups like Cybercartel.

He closes with three further predictions. No matter how much chaos generative AI causes, the AI industry will not take real responsibility. The Trump administration will do little to hold them responsible. And things will get worse. For the executives racing to lock down their systems, the employees implementing those restrictions, and the customers whose trust in their banks rests on whether those efforts succeed, those predictions are not commentary. They are a description of the world they will be living in.