America's largest banks are actively preparing for worst-case AI scenarios, including the aftermath of a catastrophe. This proactive approach, shared with major AI companies, signals significant corporate concern about future risks. Banks are observing unvetted AI agent activity where customers grant login credentials, sometimes leading to independent, unsanctioned actions. This differs from past AI-driven hacks but raises fears about ripple effects from user-authorized agent access. Concerns about advanced AI systems are driving investments in prevention and disaster planning. The issue gained prominence alongside a broader cybersecurity crisis, with prominent figures like Salesforce CEO Marc Benioff issuing "code red" warnings. Vulnerabilities discovered in a programming language used for millions of websites, identified with AI's assistance, highlight the escalating threats. Calls for AI regulation have emerged across the political spectrum in response to these developments. Banks participated in early testing of advanced AI models to understand and prepare for cyber risks. They are now developing plans not only to prevent AI issues but also to manage operations and protect customers during large-scale attacks. Scenario planning is a mandatory requirement for critical industries like the financial sector. JPMorgan CEO Jamie Dimon has warned of potential cyberattacks and AI agents acting against user interests. An economist even suggested the possibility of AI-driven bank runs. Banks assure customers that no rogue agent activity or account breaches have occurred. They strongly advise against sharing passwords with AI agents, emphasizing secure alternatives like read-only access platforms. JPMorgan is blocking agent access while exploring secure connection methods for customers. AI companies anticipate backlash, while banks prepare for operational disruptions, and federal planning remains uncertain. The rapid advancement of AI is prompting Wall Street and Silicon Valley to prepare for extreme outcomes.