🔍 Read the full analysis: Safer Financial Tasks: What Your AI Agent Needs And Doesn’t 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 their banking usernames and passwords with AI tools, saying customers are responsible for activity resulting from that access. The warning comes as financial companies discuss agent payments and trading, while safer approaches include limited access, spending caps and human approval.
Bank of America has warned customers not to share their usernames or passwords with AI agents, saying they are responsible for activity resulting from credentials they provide. The warning arrives as Visa has identified trust as a barrier to agent payments and Robinhood says customers have placed more than $250 million into accounts that AI agents trade, highlighting questions about financial automation and control over customer funds.
The email, titled “Protecting your finances in the AI World,” says credentials should never be shared with another person or provided to an outside service, including an AI-enabled tool or agent acting on a customer’s behalf. It adds that customers who share their credentials are responsible for activity resulting from that access. Bank of America also said it is “excited about what AI used responsibly can deliver.” A spokesperson confirmed the email was authentic to TechCrunch, which reported on it October 27 after a customer posted an image of it on Bluesky.
The warning addresses a basic access problem: an agent signing in with a customer’s password can appear to the bank as the customer. That can make it difficult to distinguish an action the customer authorized from one they did not, and can leave a customer facing account restrictions or a dispute over who initiated a transaction. The source article also cites an October 22 Axios report, based on unnamed sources, that banks had found third-party AI agents accessing customer accounts with customers’ passwords. The report said banks were blocking some access and revising terms; those details have not been independently established here.
The article’s author recommends granting agents limited permissions instead of master credentials: read-only access, approval for each payment, a capped virtual card, or a token restricted to a merchant, amount and time period. A separate account with a limited balance is another proposed safeguard. These controls can limit potential damage, but they do not by themselves settle who is liable if an agent makes an unwanted transaction.
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 Payment Risk
The issue is whether an AI tool receives authority broad enough to move money, and who bears the cost when its actions go wrong. A password can grant access beyond a single requested task, while a bank may have limited technical evidence to distinguish an agent’s login from the account holder’s.
Agents that read material from websites, email and documents can also face prompt injection, in which text embedded in that material attempts to steer an AI system into taking unintended actions. OWASP ranks prompt injection as a leading risk for AI applications, the article says. Narrow permissions and required approvals can reduce exposure, though they cannot guarantee that an agent will behave as intended.
For customers, possible consequences include losses, disputed transactions and account lockouts. Financial firms face the challenge of supporting automation while determining whether actions taken under a customer’s login were authorized. Businesses developing agent payments need to explain permission limits and responsibility for losses before users disable approval steps.
Products Worth Considering
DJI Osmo Pocket 3 Creator Combo Vlogging Camera with Mic 2, 1 CMOS, 4K/120fps, 3-Axis Stabilization, 128GB Card
DJI Osmo Pocket 3 Filter Bundle – Vlogging Camera with 1 CMOS Sensor and 4K/120fps
Sonos Era 100 SL Compact Microphone-Free Speaker, Black
As an affiliate, we earn on qualifying purchases.
In the United States, the article discusses Regulation E, which covers certain electronic transfers. It notes that the regulation’s definition of an unauthorized transfer excludes one initiated by a person who was given access by the consumer, subject to conditions including whether the consumer later tells the financial institution that the person is no longer authorized. The article’s author reads the language as potentially relevant to an AI agent given a customer’s login, but says no regulator or court has ruled on that application, to the author’s knowledge.
For the European Union, the article points to the Payment Services Directive, or PSD2, which requires users to take reasonable steps to keep personalised security credentials safe. It says a user who acts with intent or gross negligence can bear losses from unauthorized transactions. Whether giving credentials to an AI agent meets that standard has not been tested in court, according to the author. These are legal interpretations, not established rulings or legal advice.
The source also contrasts credential sharing with Robinhood’s reported approach of using a separate account and spending limits for agent trading. Despite those controls, Robinhood says customers “assume all risk” for trades made by AI agents. That example shows that technical safeguards and contractual allocation of losses are distinct issues.
““If you do share your credentials, keep in mind that you’re responsible for activity resulting from that access.””
— Bank of America, in the customer email confirmed as authentic to TechCrunch
Where Liability Remains Unsettled
No court or regulator ruling is cited that decides how existing US or EU consumer-payment rules apply when a customer gives banking credentials to an AI agent. The source’s legal discussion is the author’s interpretation, and the effect may depend on the facts, applicable law and the account terms.
The source does not provide the full details behind Visa’s comments to investors or Robinhood’s reported agent-trading figure, including the measurement period and comparison baseline for the $250 million. It also does not establish how many customers are using such tools or how often agents have caused losses. The reported bank response to unauthorized agents relies on unnamed sources cited by Axios, rather than details in a bank announcement.
Finally, technical restrictions may limit what an agent can do, but the source does not show that any particular control prevents prompt injection or guarantees reimbursement. The specific protections available to a customer depend on the service, permissions granted and financial provider’s terms.
Check Permissions Before Connecting
Customers considering financial agents should check whether a tool requests a password or offers a narrower permission, and read the provider’s terms on disputes and losses before enabling transactions. Where available, read-only access, transaction-by-transaction approval, spending limits and dedicated funds offer more restricted alternatives to handing over a primary account login. Customers should also know how to revoke access and contact their bank if credentials have already been shared.
For companies developing agent payments, outstanding questions include how to authenticate an agent separately from its user, restrict authority to a specific task, and explain who bears losses when an action is disputed. Developments to monitor include changes to bank policies and terms, and any regulatory or court decisions addressing agent access. The source gives no timetable for either.
Key Questions
What did Bank of America tell customers about AI agents?
It told customers not to share banking usernames or passwords with another person or outside service, including an AI-enabled agent. The email says customers are responsible for activity resulting from credentials they share.
Why is giving an agent a banking password risky?
A password may let an agent access more than the specific feature or task the customer intended. The bank may also have difficulty distinguishing an agent using the customer’s login from the customer or someone who stole the credentials.
Are customers legally responsible for every transaction an AI agent makes?
The source does not establish a general legal rule for AI-agent transactions. It discusses US and EU payment provisions but says their application to an agent given a customer’s credentials has not been decided by a court or regulator, to the author’s knowledge.
What are safer alternatives to sharing a password?
Options described in the source include read-only access, approval for each payment, a virtual card with a spending limit, a token restricted by merchant, amount and time, or a separate account with limited funds. Availability depends on the service.
Does a spending limit mean the company will cover losses?
No. A spending limit is a technical restriction, not a guarantee about liability. The source says Robinhood used a separate account and spending limits while telling customers they assume all risk for AI-agent trades.
Source: ThorstenMeyerAI.com
