🔍 Read the full analysis: AI Money Tasks: Safer Ways To Give An Agent A Hand 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 warned customers not to share banking passwords with AI tools, saying customers are responsible for activity resulting from shared access. Visa cited trust as a barrier to agent payments, while Robinhood said customers had placed more than $250 million in accounts used by AI agents. The safer alternatives described in the source include limited permissions, payment approvals and capped funds, but their protections and liability rules remain unsettled.
Bank of America has warned customers not to give banking passwords to AI assistants, saying they are responsible for activity resulting from credentials they share. The warning comes as Visa identifies trust as a barrier to AI agents making payments and Robinhood says customers have put more than $250 million into accounts used by trading agents.
A customer posted the bank’s email, titled “Protecting your finances in the AI World,” on Bluesky. TechCrunch reported on it October 27, and a Bank of America spokesperson confirmed to the publication that the email was authentic. The message says usernames and passwords should not be shared with another person or provided to an outside service, including an AI-enabled tool or agent acting for the customer. It adds that customers who share credentials are “responsible for activity resulting from that access.”
The bank also said it was “excited about what AI used responsibly can deliver.” Its warning does not describe a new general prohibition on AI tools; it focuses on sharing login credentials and the responsibility the bank says follows from doing so. The source reports that Visa told investors trust is the biggest barrier to agents paying for things. Robinhood said customers had placed more than $250 million into accounts that AI agents trade, but the source does not specify the measurement period or provide a comparison baseline.
One proposed alternative to handing over a full account password is to give an agent narrow, limited access: for example, read-only permissions, required approval for each payment, a virtual card with a spending cap, or a token restricted to a particular merchant, amount and time period. A separate, limited balance can also cap potential losses. These are safety measures, not guarantees that a provider will cover losses or that a bank will accept an agent’s activity.
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
A banking password is a broad access credential, not permission for one isolated task. An agent logging in with a customer’s credentials may appear to the bank to be the customer, making it harder to distinguish intended automation from account misuse. The same access can expose more than the payment or transaction a person meant to authorize.
The question of who bears the loss matters as these tools move from answering questions to moving money. The source describes possible limits in US and EU consumer protections when a customer has supplied access credentials, while stressing that applying those rules to AI agents has not been tested in court. Banks may also block activity they cannot verify, potentially disrupting access for customers as well as stopping unwanted transactions.
Permission limits can reduce the amount an agent is able to do, but they do not settle liability. Robinhood’s use of separate accounts and spending limits, as described in the source, did not eliminate customer exposure: the company reportedly says customers “assume all risk” for trades made by AI agents. Customers therefore need to read the terms governing the specific tool and account before enabling automated activity.
Popular right now
Bose QuietComfort True Wireless Earbuds with Active Noise Cancellation, Black
Bose QuietComfort Wireless Noise-Cancelling Over-Ear Headphones, Black
Vitamix 5200 Professional-Grade Blender, 64 oz. Container, Black
As an affiliate, we earn on qualifying purchases.
Banks Respond to Agent Logins
The Bank of America email follows a broader concern about AI services accessing financial accounts through customers’ own logins. On October 22, Axios reported, citing unnamed sources, that major US banks were finding unauthorized third-party AI agents inside customer accounts. The report said agents were often signed in with the customer’s password and that banks could not always tell whether the activity reflected the customer’s wishes.
That account of banks’ response is reporting attributed to unnamed sources, rather than a public finding by regulators. The source says banks have been blocking some agents and revising terms as they confront the difficulty of separating customer-directed automation from other access. Bank of America’s confirmed email makes the credential warning public, but does not explain how the bank identifies agent activity or what enforcement steps it may take.
AI agents can also be influenced by material they read, such as webpages, emails, invoices or documents. The source identifies prompt injection—instructions hidden in content that may steer an agent—as a major risk listed by OWASP for AI applications. That makes limiting permissions relevant even when a customer trusts the agent provider: the agent may encounter content the user did not intend to authorize as an instruction.
““If you do share your credentials, keep in mind that you’re responsible for activity resulting from that access.””
— Bank of America, in the same email
Liability Rules Remain Untested
No court or regulator ruling cited in the source determines whether giving an AI agent a banking password counts as furnishing an access device under US Regulation E. The source’s author presents that connection as an interpretation of the rule, not settled legal advice. In the EU, the source points to PSD2 duties to protect security credentials, but says courts have not tested whether sharing a password with an AI agent amounts to gross negligence.
It is also unclear how often banks are seeing agent logins, how many accounts or transactions are affected, and what controls financial firms will accept. The Robinhood figure is reported as more than $250 million in customer funds placed in accounts used by agents, but the source gives no time window or baseline. It does not establish how much has been traded, whether customers have incurred losses, or how the figure was calculated.
The source does not provide a detailed technical standard for the proposed limited-access tokens, or evidence that every bank and agent provider supports them. Nor does a spending cap by itself establish who pays after an error, unauthorized transaction or dispute. Specific protections depend on the product, provider terms and applicable law.
Check Permissions Before Connecting
Customers considering an AI service for financial tasks can check whether it offers read-only access, transaction-by-transaction approval, capped virtual cards or narrowly scoped tokens instead of asking for a full banking password. They should also read the service’s terms for its treatment of losses and disputes, and confirm with their bank how third-party access is handled. A separate account or limited balance can restrict exposure, but does not guarantee reimbursement.
For banks and companies building agent payment systems, the unresolved issue is how to grant useful authority without making an agent indistinguishable from its customer. More limited credentials, clear spending boundaries and an approval step can narrow that authority. The source offers these as safer patterns, not as a settled industry standard.
No regulatory decision or new industry-wide rule is identified in the source. The next developments to watch are how banks revise customer terms and account controls, whether payment providers adopt more limited agent credentials, and whether regulators or courts clarify responsibility when an AI agent acts with access supplied by a customer.
Key Questions
What did Bank of America tell customers?
It warned customers not to share their usernames or passwords with outside services, including AI-enabled tools and agents. The bank said customers are responsible for activity resulting from credentials they share.
Does the warning mean AI agents are banned from banking?
The source describes a warning against sharing banking credentials, not a general ban on using AI for financial tasks. The bank also said it was “excited about what AI used responsibly can deliver.” How individual banks handle agent access may vary.
What could someone use instead of sharing a password?
Options described in the source include read-only permissions, approval for each payment, a capped virtual card, a limited-purpose token, or a separate account with only a restricted amount of money. Availability depends on the bank and service.
Will a bank reimburse losses caused by an AI agent?
The source does not establish a single answer. It explains that the application of US and EU payment rules to AI agents has not been tested in court, and says customers should check the terms of the bank and agent service. A technical spending limit does not by itself settle liability.
What does Robinhood’s $250 million figure mean?
The source says customers have put more than $250 million into accounts that AI agents trade. It does not state the time window, comparison baseline, amount traded or whether customers experienced losses, so the figure should not be read as a measure of agent profits or performance.
Source: ThorstenMeyerAI.com
