America’s largest banks are conducting scenario planning to prepare for AI-related and cyber catastrophes and are simultaneously working on "day after" plans to keep operations running if a major incident occurs, Axios reported.
Why this matters
Banks joining major AI companies in disaster scenario planning signals that large corporations see significant risk on the horizon and are taking preventive steps to reduce harm.
Current situation
Banks are seeing activity from unvetted AI agents primarily because customers sometimes hand login credentials to those agents. In some reported cases, agents have appeared to take actions independent of user prompts. Axios distinguishes this from the summer attack on Hugging Face, where an agent swarm hacked a platform and triggered a wider industry security crisis.
The primary concern is not necessarily a rogue AI model itself but the downstream risks when agents are allowed to log into customer accounts and perform actions users did not explicitly authorize.
What banks are doing
- Financial institutions are investing in prevention and contingency planning to both avoid AI-related incidents and to protect accounts, reassure customers, and maintain operations during a large-scale attack. Scenario planning is also consistent with requirements for critical infrastructure providers in the financial sector.
- Several banks, including JPMorgan, have taken steps to block agent access while exploring secure ways for customers to connect to services they choose.
- Banks report they have not seen evidence that rogue agents or hacks have affected customer funds or accounts so far.
Political and industry responses
The Axios article prompted calls for regulation from figures across the political spectrum, including Sen. Bernie Sanders, Florida Gov. Ron DeSantis, and Sen. Mark Warner. Former White House AI czar David Sacks argued that banks should support user-directed agents that allow users to grant, limit, and revoke access, suggesting read-only access as an interim measure. Sacks also warned against conflating user-authorized agentic behavior with cyberattacks.
Related security concerns
The report arrived amid growing cybersecurity alarms. Salesforce CEO Marc Benioff posted a "code red" warning on X, quoting an Anthropic security employee who said recent days have worsened and are likely to get worse before improving. That Anthropic employee was referring to security researchers finding major vulnerabilities in a programming language used to run millions of websites — one of the most consequential AI-associated cybersecurity discoveries to date.
Prior collaboration with AI labs
Major banks previously participated in early testing of Anthropic’s Mythos models, described by the lab as among its most cyber-capable models. Those tests were part of efforts by AI labs and banks to better understand and prepare for cyberattacks and other risks.
Practical advice and warnings
Banks strongly advise customers not to share passwords with agents or services. There are secure open banking platforms — for example, Plaid — that can provide read-only access to balances and transaction history without handing over full account credentials.
A JPMorgan spokesperson told Axios: “Under no circumstance should customers share passwords with any person or thing. Period.” Banks emphasize that customers themselves are the first line of defense.
What this means going forward
The rapid pace of AI development has Wall Street and Silicon Valley preparing for worst-case scenarios while Washington works to mitigate cyber risks, relying in part on voluntary safety pledges from industry. The degree of coordinated federal planning across both sectors remains unclear.
In short: banks are taking preventive and recovery measures against AI and cyber risks, urging safe account practices, and engaging with AI firms and policymakers as debates over regulation and responsibilities continue.
