Citi and Bank of America Restrict AI Agents From Accessing Bank Accounts

October 19, 2026 • FinTech News
Citi and Bank of America Restrict AI Agents From Accessing Bank Accounts

Citigroup and Bank of America are drawing firmer boundaries around autonomous AI tools, restricting the access that third-party AI agents have to consumer banking accounts. The two lenders join JPMorgan Chase in raising barriers to autonomous account access, according to a Financial Times report published on October 19, 2026. The moves reflect growing concern among major U.S. banks over payment liability, customer authorization, and the possibility that automated cash management tools could eventually shift billions of dollars away from low-cost bank deposits.

Why the Banks Are Tightening Access

Citigroup has said it does not allow third-party AI agents or assistants to log in to customer accounts or execute transactions directly through its digital banking platforms, framing the move as a way to protect customer accounts and information. The policy still permits some data sharing: customers can grant certain open-banking providers limited, read-only access to account information, so services can retrieve financial data without gaining permission to move money.

Bank of America is also limiting AI agents’ access to consumer deposits while it develops protocols to establish who is responsible when an AI system authorizes or initiates a payment. JPMorgan Chase has taken a similar stance, citing risks posed by external agents operating at scale.

The restrictions point to a central tension in agentic finance. AI assistants can help consumers manage bills, compare products, and monitor balances, but giving those same systems the ability to initiate transfers introduces risks that simply do not exist when an assistant only provides information. An agent that can move money may act on incorrect information, misread an instruction, or respond to manipulated data, and a resulting transfer could land with the wrong recipient or leave a customer short of cash.

Liability Remains the Core Obstacle

Responsibility for AI-authorized payments is one of the biggest barriers to wider adoption of autonomous financial transactions. Bank of America has indicated that human approval should remain part of the process for money leaving an account while banks and AI companies work out the necessary safeguards.

That raises difficult questions. If a customer tells an agent to pay a bill and the system sends the wrong amount, who bears responsibility? A customer might authorize an agent to manage routine payments without meaning to approve every single transaction, so banks must work out how such permissions get represented, limited, and revoked.

Traceability matters just as much. Bank of America has reportedly been working with agent providers to prevent situations where customers cannot determine where their money went or how a transfer was initiated, since a reliable system needs an auditable record connecting the original instruction, the agent’s actions, and the resulting payment. JPMorgan warned during its third-quarter 2026 earnings discussion about the risk of so-called rogue agents operating at scale, and it is building its own internal tools rather than opening its systems to outside agents until stronger safeguards are ready.

The Deposit Threat Behind the Caution

Security and liability are only part of the story. AI’s growing ability to optimize personal finances also poses a competitive challenge to traditional banks. Many consumers leave substantial balances sitting in checking accounts that pay little or no interest, largely because moving that money has historically required research, paperwork, and ongoing attention. AI agents could remove much of that friction by identifying idle cash, comparing deposit products, and helping customers shift eligible funds toward better returns.

In a September 27, 2026 note titled “Is an Agentic Bank Run Coming?”, Apollo Global Management chief economist Torsten Slok warned that AI assistants could eventually move household cash into accounts paying between 3.3% and 5%, compared with a national average of just 0.1% on U.S. checking accounts. The gap is substantial: on a $10,000 balance, a 0.1% rate generates roughly $10 in annual interest before taxes, while a 4% rate generates about $400. AI tools could make it far easier for consumers to spot and act on that difference rather than relying on periodic manual reviews.

For banks, a meaningful shift of deposits toward higher-yielding alternatives could raise funding costs, since low-cost deposits underpin much of traditional lending. According to the figures cited in the report, JPMorgan Chase, Bank of America, and Wells Fargo together hold roughly $1.6 trillion in non-interest-bearing deposits, about 16% of their combined liabilities. Paying a 3% annual rate on those balances would cost the three banks an estimated $47 billion a year, close to half their combined annual earnings, though those figures illustrate a hypothetical shift in funding costs rather than a forecast that the banks will actually lose those deposits. Customers still need transaction accounts for everyday payments, and the highest advertised savings rates often come with eligibility requirements, but even so, AI could lower the effort needed to compare products and pressure banks to sharpen their offerings.

Meta’s Muse and the Push Toward AI Money Management

The debate has intensified as technology companies roll out AI assistants built to handle more personal tasks. The report links the banks’ caution to Meta’s launch of its Muse personal AI assistant in September 2026, part of a broader shift toward systems that manage activities rather than simply answer questions. In personal finance, that could mean monitoring spending, identifying surplus cash, and helping consumers manage routine financial decisions with less manual effort.

The usefulness of such tools depends heavily on the permissions they receive. An assistant that only recommends a higher-yielding account operates very differently from one that can open accounts or move funds automatically, and each added capability raises fresh questions about consent and accountability. The banks’ restrictions highlight how difficult it still is to connect consumer-facing AI products to established financial infrastructure, since technology companies want a seamless experience while banks remain responsible for protecting account access.

Regulators Take Notice

The growing restrictions have drawn attention from regulators too. According to the Financial Times, Wall Street executives have contacted the Federal Reserve about the rapid growth in consumer use of AI agents for financial transactions. The Fed declined to comment, though regulators were reported to have privately welcomed the banks’ restrictions. Officials are reportedly weighing whether greater legal clarity is needed around how consumers authorize autonomous systems to transact and who should be liable when those systems make mistakes.

Any future framework will need to balance innovation against meaningful consumer protection: overly restrictive rules could make useful automation harder to build, while weak safeguards could leave consumers uncertain about their rights when a transaction goes wrong. Clear authorization standards, accessible records, and effective dispute-resolution procedures could help build trust, and banks will also need to make sure customers understand what an agent can access. The reported discussions do not establish that the Fed or any other regulator has adopted a formal rule governing AI agents’ access to bank accounts.

How Banks May Respond to the Competitive Pressure

Restricting external agents may reduce immediate operational risk, but it does not remove the competitive pressure created by AI-driven financial optimization. Banks may need to make their own digital services more attractive to consumers who increasingly expect technology to surface better financial outcomes. One response could involve more competitive deposit products, while another could be stronger loyalty programs tying deposit balances to borrowing rates and other benefits. A Simon-Kucher senior partner cited in the report argued that banks will need to give customers compelling reasons to keep their savings with them, including more attractive borrowing terms.

Banks could also build their own AI-powered financial management tools within controlled environments, letting customers monitor balances and automate selected tasks under the bank’s own security and authorization processes. That approach would let banks participate in agentic finance without handing external systems unrestricted account access, though customer trust will remain essential regardless of who builds the tool.

What Comes Next

The decisions by Citi, Bank of America, and JPMorgan Chase highlight a defining challenge for the next phase of financial technology: autonomous AI has to operate within systems built to protect money, identity, and customer rights. The technology’s potential is real, since AI agents could help consumers manage bills, compare rates, and make routine financial decisions with far less manual effort, and could also make it easier to switch providers, sharpening competition across banking and fintech.

But those benefits depend on a trustworthy framework for execution. Financial institutions need confidence that agents are acting within authorized limits, customers need to understand what they have permitted, and everyone involved needs a workable process for resolving disputes. The deposit question adds another layer, since if AI tools make it meaningfully easier to chase better returns, banks may face pressure to compete harder for deposits even without anything resembling a systemic bank run. For now, the restrictions from Citi, Bank of America, and JPMorgan show that access to financial data and authority to move funds are being treated as separate questions, a distinction likely to remain central as banks, technology companies, and regulators work out the rules for agentic finance.