If you give an outside AI agent your bank login and it sends money where you did not mean it to, Citi’s updated online terms say Citi will treat the agent’s actions as authorized by you. Wells Fargo’s own notice describes similar terms, and TechCrunch reported that four more big banks, Citizens, PNC, JPMorgan Chase and Bank of America, moved in a similar direction.
Agents can be tricked by hidden instructions, as our explainer on prompt injection sets out, and researchers have shown a single webpage pulling a Muse user’s chats. The bank terms decide who carries the cost when that ends in a payment.
Federal rules still matter. Under Regulation E, a transfer made by someone you gave your access device to is generally not “unauthorized” unless you have told the bank that person’s access has ended. We also explained how AI shopping agents pay.
Business Pill · The Model and the Harness
An evergreen Business Engineer explainer. It teaches the general idea only and does not discuss this report.
The key insight: Citi’s new Section H says it will treat actions by a customer’s outside AI agent as authorized by the customer; Wells Fargo’s updated terms start 24 November, and TechCrunch reports four more big banks moved in a similar direction.
What Citi’s Section H Says
Citi added a Section H on the use of third-party artificial intelligence agents to its Online User Agreement in November. It says customers “are cautioned against using any third-party AI Agent” to access their accounts.
“If you disregard this caution and provide your Online Credentials to any third-party AI Agent … Citi will consider any access and actions taken by the agent as authorized by You,” the terms say. The same applies if those credentials are “further shared, leaked, or otherwise distributed.”
The terms add that customers are responsible for an agent’s “incorrect interpretation or execution of your instructions, unintended actions,” or actions that exceed the authority granted, “except as otherwise permitted under applicable law.” They do not apply to agents Citi itself provides.
A Citi spokesperson told TechCrunch that customers who grant third-party AI agents account access are responsible for activities conducted through those agents, and that “we want customers to understand the risks associated with sharing banking credentials.”

The Other Banks
Wells Fargo says its updated Online Access Agreement takes effect on 24 November. Requests sent through a third-party AI tool “may be treated as if they came directly from you,” its notice says, and it may restrict or block certain tools. American Banker reported that Wells Fargo appears to be the first bank to require customers to prove they have turned off an agent’s access.
PNC changed its terms on 18 November, Citizens emailed customers that week, and JPMorgan Chase and Bank of America announced similar policies in October, TechCrunch reported. We did not read those four banks’ terms ourselves.
Where Federal Law Comes In
Regulation E limits what a consumer pays for an unauthorized electronic transfer: the lesser of $50 or the amount taken if the bank is told within two business days, and up to $500 after that. The $500 is 10 times the $50, by our arithmetic.
The catch is the definition. Regulation E does not treat as unauthorized a transfer made by a person the consumer gave the access device to, unless the consumer has told the bank that person’s access has ended. Whether that covers an AI agent given a password is not settled in the sources we read.
The Structural Read
As we read it, Citi’s and Wells Fargo’s wording places the risk of agent mistakes on the customer who chose the agent.
As we read it, Regulation E’s definition of an unauthorized transfer is where this will be tested.
Citi Online User Agreement, Section H, November 2026
“Citi will consider any access and actions taken by the agent as authorized by You.”
Three Implications
FOR CUSTOMERS Citi says it will treat an outside agent’s actions with your login as authorized by you.
FOR AGENT BUILDERS Wells Fargo says it may restrict or block certain third-party AI tools.
FOR REGULATORS Regulation E sets $50 and $500 limits for unauthorized transfers, and its definition decides what counts as unauthorized.
What Is Not Established
No regulator has said how Regulation E applies to agent-initiated transfers in the sources we read, and none of the banks has published a case in which a customer was held liable. Citi’s own wording keeps an exception for what “applicable law” permits.
The Bottom Line
Citi says it will treat what an outside AI agent does with your login as authorized by you, and Wells Fargo says agent requests may be treated as coming from you; TechCrunch reports four more banks moved in a similar direction. What to watch is Wells Fargo’s 24 November start date and whether the CFPB says how Regulation E applies to agents.
97,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. Citi’s Section H is quoted from its Online User Agreement and Wells Fargo’s from its own notice; the PNC, Citizens, JPMorgan Chase and Bank of America policies are as reported by TechCrunch, and we did not read those banks’ terms. Regulation E is quoted from the CFPB’s published text. The 10x ratio is our arithmetic. This is not legal advice.
Sources: Citi Online User Agreement, Section H, updated November 2026 · Wells Fargo, We Are Updating Our Online Access Agreement (effective 24 Nov 2026) · TechCrunch, Tim Fernholz, 18 Nov 2026 · American Banker, Penny Crosman, 2 Nov 2026 · CFPB, Regulation E, 12 CFR 1005.2(m) · CFPB, Regulation E, 12 CFR 1005.6(b) · FourWeekMBA explainer on prompt injection, 22 Nov 2026 · FourWeekMBA explainer on how AI shopping agents pay, 21 Nov 2026 · FourWeekMBA on the Meta Muse webpage hijack, 21 Nov 2026









