🔍 Read the full analysis: How To Set Up An AI Agent For Money Tasks Carefully on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Bank of America has warned customers not to share banking credentials with AI tools, saying customers are responsible for activity resulting from that access. The warning comes as banks, payment companies and trading platforms contend with AI agents that can access money. Safer approaches limit what an agent can see or spend, but legal responsibility and protections remain unsettled.
Bank of America warned customers not to give their usernames and passwords to AI assistants or agents, saying they are responsible for activity resulting from shared credentials. The email, confirmed as authentic by a bank spokesperson to TechCrunch, highlights a growing concern as AI agents begin interacting with financial accounts and payments.
The email, titled “Protecting your finances in the AI World,” says banking credentials “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.” It adds that if customers share credentials, “you’re responsible for activity resulting from that access.” Bank of America also said it is “excited about what AI responsibly used can deliver,” according to the email quoted in the source material.
The warning matters because an agent that signs in with a customer’s password may have broad account access and can appear to the bank as the customer. That can make it difficult to distinguish an action the customer authorized from one caused by an error, misuse or compromised agent. The source article says Axios reported on October 22, citing unnamed sources, that major U.S. banks were finding third-party AI agents inside customer accounts, often using customers’ own passwords. That report does not establish how widespread the activity is.
The source also describes different approaches to agent access: read-only permissions, requiring approval for each payment, limited virtual cards, tokens restricted by merchant, amount and time, or a separate account with a capped balance. These are risk-reduction options, not a guarantee against losses. Robinhood was reported to have said customers had placed more than $250 million into accounts that AI agents trade; the source says the company also tells customers they “assume all risk” for agent trades.
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
Why Agent Access Changes Account Risk
A banking password is not a narrowly scoped permission. It may allow an agent to do many of the same things as the account holder, while leaving the bank with limited ability to identify which actions were intended. One credential can grant broad access, even when the requested task is small, such as paying a particular bill.
There is also a potential gap between technical safeguards and legal responsibility. The source article interprets U.S. and EU rules as raising questions about whether consumers who give an agent access could bear losses from its actions. It explicitly says the application of those rules to AI agents has not been tested by a regulator or court, as far as the author could find. Readers should not treat that interpretation as a settled legal ruling or legal advice.
For consumers, the practical issue is not only whether an agent can make a mistake. It is also what happens if it makes an unintended payment, follows misleading instructions, or triggers a bank security lock. Limiting access and requiring approval can reduce the amount an agent can do without permission, though the available controls depend on the bank, payment service or agent provider.
Popular right now
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
Ninja CREAMi 7-in-1 Ice Cream Maker, NC301
As an affiliate, we earn on qualifying purchases.
Banks Respond to Financial Agents
The warning follows wider discussion about AI agents operating in customer accounts. According to the source material, Axios reported on October 22 that banks were encountering third-party agents logging in with customer credentials and were reviewing how to respond. The report cited unnamed sources; the underlying scale and particular incidents are not established in the material provided.
On the same Tuesday as the reported Bank of America email, Visa told investors that trust was the largest barrier to agents paying for goods and services, according to the source. Robinhood’s reported agent-trading accounts show that some customers are already placing money in systems that let agents trade. These developments point to a practical distinction: an agent may be able to act, but customers still need to know what it can do, what approval it needs and who is responsible if something goes wrong.
The source identifies prompt injection as another concern: instructions embedded in web pages, emails or files may influence an agent that reads them. It says OWASP lists prompt injection as a leading risk for AI applications. This does not mean every agent will be manipulated, but it underlines why financial permissions should be limited rather than based on trust in the agent alone.
“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 an email to customers
Liability Rules Remain Untested
No court or regulator ruling is cited in the source material that decides how U.S. electronic-transfer protections apply when a customer gives an AI agent account access. The author’s reading of Regulation E is an interpretation, not an established legal conclusion. The same is true of whether sharing credentials with an agent would count as a failure to protect security details under EU payment rules.
The precise terms of coverage and responsibility may vary by financial institution, product and circumstances. The source material does not provide the full text of the Bank of America email beyond the quoted passages, nor does it establish how many customers have connected agents to accounts or how often those agents have caused losses. It is also unclear what standards platforms will use to approve, monitor or restrict agents.
Check Permissions Before Connecting
Before connecting an agent to a financial account, customers can review the provider’s terms and the bank’s policies, and check whether access can be restricted to viewing balances or a specific task. Where available, payment-by-payment approval, spending caps, merchant limits and separate low-balance accounts can reduce the potential impact of an error. A customer should avoid giving an agent a primary banking password when a narrower permission is available.
Companies building agent payment systems will need to explain what an agent can access, how customers approve transactions, how to revoke access and who bears losses. Bank warnings, payment-industry comments and platform terms are evolving; further guidance or regulatory decisions may clarify the responsibilities. Until then, customers should treat broad account access as a financial risk, not as a routine setup step.
Key Questions
The bank says credentials should not be shared with outside services, including AI tools or agents. An agent using a customer’s login may have broad access and can be difficult for the bank to distinguish from the customer.
Can a customer lose fraud protection after giving an agent access?
That is not settled for AI agents. The source article discusses U.S. and EU rules that could raise questions about responsibility, but says their application to AI-agent access has not been tested in court or decided by a regulator. Check the relevant account terms and seek qualified advice for a specific dispute.
What is a safer way to let an agent handle a payment?
Use the narrowest available permission, such as read-only access, approval for each payment, or a virtual card or token limited by merchant, amount and time. A separate account with limited funds can also restrict potential losses.
What is prompt injection, and why does it matter for money tasks?
Prompt injection refers to instructions placed in material an agent reads, such as a web page or email, that may influence its actions. For financial tasks, limited permissions and human approval can help contain the consequences if an agent is misdirected.
Source: ThorstenMeyerAI.com
