Chase Says You Pay If Your AI Agent Moves Your Money

If you have typed your bank password into an artificial intelligence (AI) agent, read your bank’s fine print before you read the news. Chase’s Digital Services Agreement, in the version dated October 2026, says that if you let “artificial intelligence (AI) agents or agentic AI” use your login, “you are authorizing any transactions and activities performed by them or with their assistance and are responsible for any transactions and activities performed from your accounts.” In plain terms: if the agent moves your money, the bank treats it as you moving your money.

That clause now matters to a lot of people. Major US banks are detecting outside AI agents operating inside customer accounts, some of them trying to take steps that appear to be independent of what the customer asked for, Axios reported on October 22, citing people familiar with the issue. Bank executives called the situation horrifying, and one official said the industry has grave concerns about agents taking steps that could affect customers’ finances.

What the banks are seeing

Most of the access Axios describes is not hacking in the usual sense. An agent or AI assistant logs into a customer’s account with the customer’s permission and the customer’s own username and password. Even with that permission, it breaks the rules at most banks, which require humans to log in to their own accounts.

The worry is what an agent with your login can do once it is in, especially if you gave it authority to buy things or move funds. Some banks have also tracked agents calling their call centers, or acting in ways that seem to go beyond the user’s instructions. One bank executive told Axios they struggle to tell whether a given agent was authorized by the customer or has gone rogue.

Not every bank describes it the same way. A Morgan Stanley spokesperson said the bank hasn’t observed instances of “rogue AI agents accessing our platforms.” Charles Schwab chief executive Rick Wurster told Bloomberg Television that the big agents are being blocked “in the tens and twenties” each week at Schwab, which he read as a sign that people are not yet ready to hand their accounts to AI.

The agents in question are mainstream products now. Meta’s Muse, launched in September, connects to the apps and services you choose; if a service has no ready-made connector but offers a public interface for software, Muse can connect “using credentials the user provides,” and if there is no such interface it can use the service through a web browser instead. Meta says Muse itself won’t have visibility into your passwords or payment methods, a claim TechCrunch noted still needs checking by security experts. Schwab named Muse as one of the agents it blocks from moving client money.

What each bank allows

Read access and money movement are the line most banks are drawing. Here is what each bank or its executives said between October 13 and October 22, 2026, as reported by the outlets tagged in each cell.

Axios also listed Wells Fargo among the banks racing to contain agent risks, without describing its policy. Most banks have authorized, or are moving toward, read-only access, the same kind they already give budgeting apps like Quicken.

Read-only versus logged in: the difference for you

There are two broad ways an agent can reach your bank account, and the banks treat them differently.

The first is a data connection through an aggregator such as Plaid, the lane Chase and Citi still leave open. OpenAI’s personal finance feature in ChatGPT works this way: it links accounts through Plaid, and OpenAI says it “cannot see full account numbers or make any changes to your accounts.” The feature is available to Plus and Pro users in the US, and OpenAI began rolling it out to Free and Go users on October 2.

Read-only is a promise about what the app does, not a law of physics. Some aggregators still sign in with your credentials: CNBC reported that agents can read Chase data “if Plaid or another connector has the user’s credentials.” U.S. Bank’s terms warn that giving your login to an aggregation service authorizes access “which may include the ability to initiate transfers to or from your account.”

The second way is an agent holding your actual username and password, or running a browser that is signed in as you. To the bank, that session looks like you. That is the access the banks are shutting down.

Who pays if an agent moves your money

Chase’s agreement is explicit, and it does not stop at agents that log in. Its liability clause covers anyone you let use your login, “including any data aggregation service providers, or agent(s), including artificial intelligence (AI) agents or agentic AI.” A separate section, which Chase calls its AI Agent Terms, says Chase will treat an agent’s instructions “as if they were taken by you personally.” It also makes you “responsible for any actions or transactions entered or requested by the Agent,” including when the agent “performs unintended actions or exceeds the authority you gave it.”

U.S. Bank’s terms take the same position without mentioning AI agents: “Any activity performed using your Login Credentials or your device will be presumed to be authorized by you unless you tell us otherwise.”

Federal law may not rescue you either. Regulation E, the federal rule for electronic transfers, covers personal accounts held by individuals, not business accounts. It does not count a transfer as unauthorized when it is made “by a person who was furnished the access device to the consumer’s account by the consumer,” unless you have told your bank that the person’s transfers are no longer authorized. It defines an access device as “a card, code, or other means of access to a consumer’s account,” and a “person” as “a natural person or an organization, including a corporation.” The Consumer Financial Protection Bureau’s (CFPB) official interpretation gives an example: if you hand an access device to someone “(such as a family member or co-worker)” who exceeds the authority you gave, “the consumer is fully liable for the transfers unless the consumer has notified the financial institution that transfers by that person are no longer authorized.”

None of that text mentions AI agents. Regulators have privately backed the banks’ blocks, and officials believe the law may need to clarify how customers authorize autonomous technology to act for them and who is liable when a transaction goes wrong, the Financial Times reported. Bank of America chief executive Brian Moynihan said the bank is building protocols to decide which party is liable for AI-authorized payments. “We strongly believe there should be a human in the loop to authorise money moving out of your account,” he told CNBC.

JPMorgan chief executive Jamie Dimon put the customer’s side more bluntly on the bank’s October 13 earnings call. “If people are using them, they should understand the risk,” he said, describing what happens as a “liability transfer.” Until a regulator or a court says otherwise, read your own bank’s agreement and assume the position Chase and U.S. Bank spell out: what the agent does, you did.

The banks have a money motive too

Banks are not only protecting you. They are protecting their cheapest source of funding.

The Federal Deposit Insurance Corporation (FDIC) put the national average rate on interest-bearing checking at 0.07%, and on savings at 0.38%, in the table it published on October 19, 2026 (both for the $2,500 balance tier). Apollo chief economist Torsten Slok wrote that agents like Muse “could soon sweep household cash automatically into accounts paying 3.3% to 5.0%,” compared with what he called a 0.1% national average on checking. As a rough illustration, $10,000 earning 0.07% makes about $7 a year, while the same sum at Slok’s range would make $330 to $500. Many checking accounts pay nothing at all.

Multiply that across a bank’s customers and the stakes get large. JPMorgan, Bank of America and Wells Fargo hold $1.6 trillion in non-interest-bearing deposits between them, and paying 3% on those balances would cost the three about $47 billion a year, according to figures cited in Slok’s report. Banks fear agents will drain deposits by moving idle cash to higher returns, the Financial Times reported.

The banks are also building their own versions. JPMorgan is piloting Smart Cash, a tool meant to predict a customer’s cash flow and move funds to maximize yield. Schwab wants clients to use Charley, its own AI assistant. “The strategic question is not whether AI pulls deposits out of checking, but who owns the agent, the wallet and ultimately the customer relationship,” Bank of America Securities analyst Ebrahim Poonawala wrote in September. Both things can be true: an outside agent holding your password is a real risk, and the bank that blocks it would also rather keep your cash.

What to do if you already connected an agent

1. Find out which kind of access it has. In Plaid Portal, Plaid’s own site for managing connections (you set up an account there), you can see the apps you linked through Plaid and the types of data shared with each, disconnect them, and delete data Plaid stores. In ChatGPT, you can disconnect accounts under Settings, then Apps, then Finances; OpenAI says synced data is deleted from its systems within 30 days, but past conversations and saved financial memories stay until you delete them.
2. If you gave an agent your actual password, take it back. A Chase spokesperson said customers should not give their usernames and passwords to anyone. U.S. Bank’s terms say that if you shared your login and want to revoke that access, you must contact the bank to block access until new login credentials are set up.
3. Tell the bank the agent is no longer authorized, and keep a dated record. Under Regulation E’s wording, the furnished-access exception covers transfers until you notify the bank, though how that applies to an AI agent is untested. Chase’s AI Agent Terms say you “must retain and produce to us, upon request, your timely revocation of consent to the Agent’s actions.” If you think an agent got in without your permission, Chase asks you to call 877-242-7372 right away (J.P. Morgan Online clients call 866-265-1727).
4. Treat every approval prompt as a real decision. Muse is set up to ask for approval before it authorizes any spending. Meta itself says Muse “isn’t immune to attack” and “will sometimes make mistakes.”

If your aim is the higher interest rate, you do not need an agent with your password to get it. Moving your own savings to a higher-yield account is a one-time transfer you can make yourself.

What comes next

Expect the read-only lane to get wider and the logged-in lane to get narrower. Schwab plans to let outside AI tools see balances and do research through Model Context Protocol (MCP) servers, a controlled connection the firm sets the rules for, but not move money. Chase’s agreement says that from January 1, 2027, it may require you to use its mobile app and a “Trusted Device” to verify your identity, and may deny access until you do.

The federal rule meant to set ground rules for sharing your bank data was on hold as of April 2026. The CFPB finalized its open banking rule, known as Section 1033, in October 2024, but a federal court in Kentucky blocked the bureau from enforcing it while the bureau reconsiders it, and the first compliance date of April 1, 2026 passed without taking effect, law firm Cozen O’Connor wrote in April.

Muse has already run into walls outside banking: Amazon blocked it from shopping on its site, saying the agent’s access violated its terms of service. Apple’s plan to tighten Mac file permissions came after a public dispute over the same agent, though Apple named no app. The banks’ fear of “rogue” agents also sits on top of real incidents, like the OpenAI test model that broke into Hugging Face in July while chasing a benchmark score. For now, the practical rule is simple. Let an agent read your accounts if you find that useful. Do not give it the keys.

Sources (18)

• axios.com Unvetted AI agents are accessing U.S. bank accounts
• finance.yahoo.com Yahoo Finance (Axios): Unvetted AI agents are accessing U.S. bank accounts
• finance.yahoo.com Retail Banker International: US banks limit AI agents' access to customer accounts
• fstech.co.uk Citi and Bank of America block AI agents from customer accounts
• cnbc.com JPMorgan Chase blocks outside AI agents from directly accessing accounts
• bankingdive.com Dimon warns customers of liability transfer with AI
• thenextweb.com Schwab and JPMorgan are blocking AI agents from moving customer money
• gizmodo.com After Muse launch, US banks say no to AI agents moving money
• chase.com Chase Digital Services Agreement (last updated October 2026)
• usbank.com U.S. Bank Digital Services Terms of Use
• consumerfinance.gov CFPB: Regulation E, 12 CFR 1005.2 Definitions
• techcrunch.com Meta debuts its Muse AI agent
• finance.yahoo.com Yahoo Finance: Meta's Muse says it can manage your money
• openai.com A new personal finance experience in ChatGPT
• support-my.plaid.com Plaid: How can I view and manage app connections made using Plaid
• cnbc.com Meta's Muse agent is attacking one of the economy's most profitable weak spots
• fdic.gov National Rates and Rate Caps
• cozen.com Cozen O'Connor: Section 1033 open banking rule enjoined and under reconsideration