Major US Banks Restrict Third-Party AI Access to Customer Accounts

Four major U.S. banks, including Citi, JPMorgan Chase, Bank of America, and Charles Schwab, have announced or implemented restrictions on third-party AI agents accessing customer accounts, citing concerns over potential fund mismanagement and security risks[1]. This action marks a cautious approach to integrating autonomous AI tools into financial services.

This shift directly impacts how consumers can manage their money with AI tools, as their chosen applications may face limitations or be blocked from interacting with bank accounts[1][9][10]. The financial sector is clarifying liability and data security boundaries in response to evolving AI capabilities[6][7].

What the records show

Citi recently announced restrictions to protect customer security and prevent fund mismanagement by AI agents[1]. This follows a trend among other large U.S. banks. JPMorgan Chase CEO Jamie Dimon voiced concerns about "rogue agents" and "liability transfer"[6][7]. Dimon stated that customers should understand the risks of allowing AI agents to access their information[6]. He also confirmed JPMorgan would not permit unvetted outside AI agents into its systems[7][8]. The bank plans to build its own tools to protect customers from AI agent risks[7][9].

Bank of America CEO Brian Moynihan disclosed the bank hired 1,000 external staff for "Mythos remediation"[11][12]. This remediation addresses software vulnerabilities exposed by powerful AI models, specifically Anthropic's Mythos, which was released to financial institutions in April 2026[11]. These contractors joined approximately 3,000 internal employees already working on repairs, with the total cost for this remediation and compliance spending amounting to "a couple hundred million dollars plus" for the quarter[11][12][14].

Charles Schwab CEO Rick Wurster indicated the firm would block outside AI agents from moving client money[10]. Schwab plans to use Model Context Protocol (MCP) servers to allow client-chosen AI models to check balances and conduct research under the firm's rules, but without transaction capabilities[10].

What changed and for whom

The financial industry is adopting a more defensive stance against potentially disruptive AI agents[1]. This directly affects customers who use, or plan to use, AI agents for financial management[1][9][10]. Banks are emphasizing security and liability concerns over extensive AI integration, clarifying where they draw the line on external access[6][7][9]. For financial institutions, this means significant investment in cybersecurity and internal AI development, alongside managing external AI threats[8][11].

What the subject has said publicly

JPMorgan Chase CEO Jamie Dimon cautioned that allowing AI agents access to bank accounts involves a "liability transfer"[7]. He stated, "If data is being misused, they should understand it's being misused"[7]. Dimon emphasized the bank's commitment to protecting clients, stating, "we're not going to allow just any outside agent to come in inside the company because it's too risky for the clients and for JPMorgan"[7][9]. Jeremy Barnum, JPMorgan's chief financial officer, added that the bank has a history of opposing credential sharing by third parties, noting, "It's not secure"[9].

Bank of America CEO Brian Moynihan explained that "Mythos remediation" was a response to vulnerabilities identified by advanced AI models, necessitating external experts to patch software flaws[11][12]. Charles Schwab CEO Rick Wurster affirmed the firm's intent to protect client assets, stating, "We're not going to allow them to move money" when referring to outside AI agents[10].

What we could not confirm

We could not confirm the specific long-term impact on financial innovation within U.S. banks due to these restrictions. The total aggregate cost of AI-related software remediation efforts across all major U.S. banks, beyond Bank of America's stated figures, remains unconfirmed in the provided records.