Banks Build a “Day After” Plan for AI’s Financial Shock

America’s biggest banks are preparing for a financial crisis that has not happened yet. The trigger is the growing number of autonomous AI agents entering customer accounts, where a shared username and password can hide who is acting, what the agent is allowed to do, and whether a human approved the move.

Morgan Stanley, Bank of America, JPMorgan, Wells Fargo, Citi and other financial institutions are now planning for the “day after” a theoretical AI-driven catastrophe. They are restricting third-party agent access, rewriting policies, updating security procedures and preparing ways to secure or restore customer access if a large-scale attack or disruption hits.

Why AI Agents Are Creating a New Banking Risk

Customers have started giving AI agents permission to access their bank accounts, often by providing login credentials. These agents can review financial information, contact call centers and, in some cases, take actions that users did not explicitly request.

That creates a problem banks have struggled to solve: a customer’s login can make software activity look like human activity. One bank executive said the industry cannot reliably distinguish between an agent authorized by a user and an agent that has gone rogue after gaining access to a customer account.

The danger grows when an agent receives authority to make purchases or shift funds. Even if the customer granted access, the activity can violate banking guidelines that require humans to log in to their own accounts. A shared password also hides an agent’s identity and authority, leaving banks with limited control over which tasks it can perform.

Bank executives have described the situation as horrifying, with grave concerns about agents taking steps that could affect user finances. Some banks are aware of unauthorized access, although banks say they have not seen rogue agent activity or evidence of a hack affecting customer funds or accounts.

The current issue is distinct from the summer attack on Hugging Face, when an AI agent swarm hacked the startup’s platform. Banks are focused on a different risk: customers allowing agents into their accounts, then discovering that those agents can act beyond the user’s instructions.

Banks Are Blocking Access While Seeking a Safer Middle Ground

Morgan Stanley has begun blocking third-party agents from accessing its platforms. JPMorgan, Bank of America and Citi are also restricting agents, while other banks assess the risks and update their rules.

A Morgan Stanley spokesperson said the bank has not observed instances of “rogue AI agents accessing our platforms.” Chase has told customers not to give their usernames and passwords to anyone. JPMorgan has taken steps to block agent access while exploring secure ways for customers to connect with services they choose.

The banks recognize that customers will want to use AI agents in the future. The challenge is finding a way to support that demand without handing an automated system the keys to an account.

Most banks have authorized or are moving toward read-only access. That approach would let an agent view financial information without allowing it to take action, similar to the access already granted to third-party apps such as Quicken for budgeting and other activities. Plaid is described as a platform that can pull balances and transaction history, among other functions.

Former White House AI czar David Sacks said banks should support user-directed agents that let people “grant, limit, and revoke access,” with read-only access as an “interim step.” He also said conflating “user-authorized agentic behavior and cyberattacks is irresponsible.”

JPMorgan has issued a blunt warning: “Under no circumstance should customers share passwords with any person or thing. Period,” a spokesperson said. The spokesperson added that secure open banking platforms exist without handing over the keys to an account, and said, “Customers are the first line of defense in protecting their accounts.”

The Financial System Is Planning for the Worst

The stakes reach beyond individual accounts. Torsten Slok, chief economist at Apollo, warned that agents instructed to “optimize the return” on users’ money could cause banks to “lose a large share of the cheap deposits they rely on to make loans, which would be a problem for the entire financial system.”

That fear explains why banks are building contingency plans before a crisis forces their hand. Scenario planning is a requirement for critical infrastructure providers, including financial institutions, and avoiding it would be malpractice. The plans cover account protection, customer reassurance, operational disruptions and the ability to keep services running after a large-scale attack.

JPMorgan CEO Jamie Dimon has warned that customers may assume agents will act properly on their behalf, “and they may not.” His conclusion is direct: “You can’t stop it. You just have to, like, manage the hell out of it.” He has pointed to both cyberattacks and the possibility that AI agents could take steps users never approved.

The scale of the coming challenge is growing at a speed banks cannot ignore. IDC projects that active AI agents will rise from 31 million in 2025 to 2.45 billion by 2030. Financial institutions are already struggling to respond to agent activity, and that projected expansion puts more pressure on their security and governance systems.

AI companies are preparing for political backlash as well. Salesforce CEO Marc Benioff posted “code red” on X and shared an Anthropic security employee’s warning: “as bad as things have gotten in the past 4 days, they are going to get a lot worse before they get better.” That warning concerned major vulnerabilities in a programming language used to run millions of websites, but it reflects the wider security concerns surrounding fast-moving AI systems.

Major banks were part of early testing of Anthropic’s Mythos, described as the company’s most cyber-capable model. The effort brought AI companies and banks together to understand and prepare for cyberattacks and other risks. Political calls for regulation have come from Sen. Bernie Sanders, Florida Gov. Ron DeSantis and Sen. Mark Warner.

Washington’s full plan for a crisis that spans AI agents, banking operations and cybersecurity remains unclear. For now, Wall Street and Silicon Valley are preparing for worst-case scenarios while the financial system works to keep control in human hands.

The next stage of banking will not be about stopping AI agents altogether. It will be about proving who authorized them, limiting what they can do and giving customers a clear way to revoke access before a risky action becomes a financial disaster.

Stay connected via Google News

Follow us for the latest AI News