🔍 Read the full analysis: How To Keep An AI Agent’s Money Access On A Short Leash on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Bank of America has warned customers not to share banking passwords with AI tools, saying customers are responsible for activity resulting from shared access. As financial firms explore agent-driven payments and trading, limited permissions, spending caps and explicit approvals can reduce exposure, but legal responsibility and platform protections remain uncertain.
Bank of America warned customers this week not to give banking usernames or passwords to AI assistants, saying customers who share credentials are responsible for activity resulting from that access. The warning comes as Visa identifies trust as a barrier to agent payments and Robinhood says customers have placed more than $250 million into accounts that AI agents trade, raising questions about the financial authority users grant to software.
Bank of America’s email, titled “Protecting your finances in the AI World,” says credentials should not be shared with another person or an outside service, including an AI-enabled tool or agent acting on a customer’s behalf. A customer posted the email on Bluesky, TechCrunch reported it on October 27, and a bank spokesperson confirmed to TechCrunch that the email was authentic. The bank also said it is “excited about what AI used responsibly can deliver.”
The warning is about more than an agent making an unwanted payment. When an AI tool logs in with a customer’s password, its activity may look to a bank like the customer’s own activity—or like a stolen login. That can complicate investigations and potentially lead to an account being restricted. An agent can also encounter instructions hidden in a web page, email or document that attempt to steer its actions, a risk known as prompt injection.
The article’s proposed alternative is to give agents narrower permissions: read-only access, approval before each payment, a virtual card with a spending cap, or a token limited to a particular merchant, amount and period. A separate account holding only money the user can afford to risk is another possible safeguard. These controls can limit exposure, but they do not by themselves settle who is liable if an agent acts improperly.
Before your agent touches money
Give it a valet key, not the master key
Logging in with your password makes the agent you. A credential made for the agent can do only the job you gave it, and you can take it back.
- What it opensEverything you can do: transfers, payees, settings, statements
- LimitsNone that the agent cannot get around by logging in as you
- What the bank seesYou, or a thief. It cannot tell which
- If a web page tricks itThe trick reaches your whole account
- To take it backChange your password everywhere it was used
- If money goes missingUS rules can exclude transfers by someone you gave access to; Bank of America: “you’re responsible”
- What it opensOnly what you allowed: read-only, one card, one pot of money
- LimitsBuilt in: amount, merchant, time window, approval on your phone
- What the bank seesAn identified agent acting for an identified person
- If a web page tricks itThe damage stops at the limit
- To take it backDisconnect the agent or delete its card. Your password never moved
- If money goes missingDepends on the terms. Read them first: some still put all risk on you
A comparison of two set-ups, not a measurement. Which limits exist depends on your bank, card and app.
Sources: Bank of America customer email, confirmed to TechCrunch (27 Oct 2026); 12 CFR 1005.2(m); OWASP LLM01. Comparison: Thorsten Meyer.
Powered by Thorsten Meyer AI
The Cost of Sharing a Banking Login
The distinction between an agent and a person authorized to use an account could matter when a transaction is disputed. In the United States, Regulation E defines an unauthorized electronic transfer in a way that excludes some transfers initiated by a person the consumer furnished with access, unless the consumer has told the financial institution that the person is no longer authorized. The article says this wording could be relevant when a customer gives an AI agent access, but no regulator or court has ruled on that application.
In the European Union, the Payment Services Directive requires users to take reasonable steps to keep personalized security credentials safe. The article says whether giving credentials to an AI agent could amount to gross negligence under the directive has not been tested in court. These are interpretations of legal text, not settled findings or legal advice. The legal treatment of credentials supplied to AI agents remains unresolved.
For businesses building agent-payment products, limited access can reduce the amount an agent is able to do if it makes a mistake or is manipulated. But a spending limit is a technical control, not a promise that a provider will cover losses. Robinhood, despite using a separate account and spending limits for agent trading, says customers “assume all risk” for trades made by AI agents. Customers need to understand both the permissions they grant and the terms governing responsibility.
Products Worth Considering
DJI Osmo Pocket 3 Creator Combo Vlogging Camera with Mic 2, 1 CMOS, 4K/120fps, 3-Axis Stabilization, 128GB Card
Logitech MX Keys S Wireless Keyboard, Graphite
DJI Osmo Pocket 4 Gimbal Camera Pro Kit, 18-Piece Bundle
As an affiliate, we earn on qualifying purchases.
Banks Confront Third-Party Agents
The warning follows a report by Axios on October 22, which cited unnamed sources who said major U.S. banks were finding unauthorized third-party AI agents inside customer accounts. The agents were reportedly logging in using customers’ own passwords. According to that report, banks often could not tell whether the activity reflected a customer’s instructions. Axios also reported that banks were blocking agents they regarded as unauthorized and revising terms.
The article places that reported bank activity alongside two signs of commercial interest: Visa told investors trust is the largest barrier to AI agents paying for goods, while Robinhood said customers had put more than $250 million into accounts traded by AI agents. Those statements indicate activity and concern, but they do not establish how many customers are involved, how the funds are performing or how frequently agents make mistakes. The central issue remains the access method: a shared password can give an agent far broader control than a single task requires.
“Your username and password “should never be shared with another person or provided to an outside service, including an AI-enabled tool or agent that offers to act on your behalf.””
— Bank of America, in its customer email
Liability Rules Remain Untested
It is not yet clear how courts or regulators would apply U.S. electronic-transfer protections to an AI agent using credentials supplied by an account holder. The article’s reading of Regulation E is an interpretation; it says no regulator or court has addressed that specific question. In the EU, it is also unresolved whether giving credentials to an agent could meet the legal standard of gross negligence.
The scope of the reported bank activity is also uncertain. Axios cited unnamed sources, and the source material gives no count of affected accounts or a detailed account of how banks distinguish authorized agent use from unauthorized access. The source does not specify the terms, performance or exact safeguards associated with the Robinhood accounts. Nor does a technical permission limit alone establish how losses would be handled.
Check Permissions Before Connecting
Customers considering an AI agent should check whether the financial service offers limited or read-only access, transaction-by-transaction approval, capped virtual cards or account-level spending limits. They should also read the provider’s terms to see who is responsible for transactions and how to revoke access. If a password has already been shared, users can change it, remove connected access where available and contact their bank about the account.
Financial institutions and agent providers will need to clarify what permissions their products use, how customers can review agent activity and what happens when an agent takes an action the customer did not intend. Until legal rulings or clearer product terms establish who bears losses, the source article’s practical advice is to grant an agent only the narrow access needed for its task and keep a person in the approval loop for payments.
Key Questions
The bank says credentials should not be given to outside services, including AI tools, and warns that customers are responsible for activity resulting from shared access. A password may also give an agent broader account control than needed for a single task.
Does a customer definitely lose fraud protection if an agent uses their login?
That is not settled in the source material. It interprets U.S. and EU rules as raising possible risks, but says the specific legal questions have not been decided by courts or regulators. The bank’s warning is not itself a court ruling on liability.
What is a safer way to let an AI agent handle payments?
Possible safeguards include read-only access, approval for each payment, a virtual card with a spending cap, or a token limited by merchant, amount and time. A separate account with limited funds can also restrict potential losses.
What is prompt injection?
Prompt injection is an attempt to steer an AI system through instructions embedded in material it reads, such as a web page, email or document. The source article identifies it as a risk for agents that interact with online content.
Do spending limits mean an agent provider will cover losses?
No. Technical limits can constrain what an agent can do, but they do not establish who is legally or contractually responsible. Robinhood, according to the source article, says customers assume all risk for trades made by its AI agents.
Source: ThorstenMeyerAI.com
