Chime Opens the Door to AI Agents. Wells Fargo Shifts the Risk to Customers.

TL;DR

Banks are splitting into two camps on AI agents banking access. Chime is building secure connections so members can bring their own agent, with permissions, while working alongside its specialist finance assistant, Jade. Wells Fargo will allow agent access in some circumstances but, from 24 November 2026, its agreement puts responsibility on customers to manage access and prove revocation. The bigger story is not just whether agents can connect, but who is accountable when they act. For fintechs, neobanks and merchants, access controls, revocation and evidence are becoming product and payments questions, not just legal fine print.

Which banks are letting customers’ AI agents in?

Chime is preparing secure connections for members’ preferred general-purpose AI agents, while JPMorganChase and Wells Fargo have been clarifying customer liability in their online-banking terms and some community banks are proceeding cautiously.

Chime CEO Chris Britt said on the company’s 4 November 2026 Q3 earnings call that members would be able to control which agents can access their accounts and what those agents can do with their money. The external agents are intended to work alongside Jade, Chime’s own personal AI finance assistant, which provides financial coaching and spending insights. Chime has not given a launch date for the connections.

That is a bring-your-own-agent proposition, but not a promise that every agent can do anything. The distinction matters: granting bank account access for AI agents means defining what they can see and which actions they can take, not simply opening a digital door.

The approach contrasts with banks’ terms-based responses. American Banker reported on 10 November that community banks were moving slowly on agent adoption, citing data security and the potential loss of human service. The market is not choosing only between adoption and rejection; it is deciding how much control and accountability to build around access.

What exactly did Wells Fargo change?

Wells Fargo’s updated Online Access Agreement, effective 24 November 2026, says using an AI agent is “AT YOUR OWN RISK” and places key responsibilities for its activity and revocation on the customer.

The agreement says Wells Fargo has no obligation to investigate or verify an agent’s authority or the validity of its instructions. It also says the bank does not monitor, screen or review agent activity, including transactions initiated by an agent. Requests and transactions made through authorised representatives or delegates, including their AI agents, may be treated as authorised unless the customer notifies the bank and the bank has had reasonable opportunity to act.

Section 15(f) adds an operational burden: customers must revoke an agent’s authorisation in a timely way, take further action to stop its activity, retain evidence of revocation and provide that evidence on request. Until the bank is notified of revocation, customers remain responsible for the agent’s activities, subject to applicable law and regulation.

Wells Fargo also reserves the right to require agent registration, change or restrict APIs, and limit, prohibit, prevent, suspend or terminate an agent’s access. In other words, the agreement does not simply ban agents. It allows for access while placing the burden of managing and documenting it largely on customers.

Who is liable when an AI agent goes rogue?

The unresolved issue is whether a customer can meaningfully prove an agent has stopped acting, and who bears responsibility if it misbehaves despite the customer’s instructions.

Douglas Borthwick, a former Morgan Stanley and Standard Chartered executive, told American Banker: “It isn't realistic to expect customers to produce one. Telling a chatbot to stop leaves nothing in the customer's hands that anyone else can check.” A screenshot may show that a person asked an agent to stop, but it cannot by itself establish that the agent’s access ended or that it ceased activity.

Even proof of revocation may not answer what happens if an agent acts on an ambiguous instruction or goes rogue. Elaine Duffus, a senior specialised consultant at Wolters Kluwer, posed the underlying question: “Who's at fault then? Is it me? Is it the bank? Is it the AI provider?” Duffus has called for an agent-governance layer with a kill switch.

This is the quieter, bigger part of the Wells Fargo change. Access rules can be adjusted, but liability language determines who may have to manage a failure after access has been granted. Agentic payments therefore need more than a permission screen: they need clear authority, usable revocation, a record of what happened and a way to contain harmful activity.

That intersects with fraud controls as well as account terms. Sardine’s Know Your Human analysis examines how agent access collides with identity and fraud systems.

What does Chime’s open-door approach say about where banking is heading?

Chime is positioning external agents as a way to help customers discover and switch financial products, while reserving complex financial tasks for trusted specialist agents such as Jade.

Britt said, “We don't believe that a single agent will be best suited to manage every area of your life.” For activities such as moving money, paying bills or managing investments, he argued, consumers need financial expertise, safeguards and accountability as well as intelligence. That is a product strategy as much as a technology position: the general-purpose agent may help a customer navigate, while a specialist agent is intended to handle finance-specific tasks.

Britt also said agents could make it easier for consumers to find better-value financial products and switch institutions, “eroding the inertia that has protected big banks for decades.” That creates a potential distribution challenge for incumbent banks and an opportunity for fintechs that can offer compelling products and controlled connections.

Chime’s performance provides commercial context for its willingness to invest. In Q3 2026, revenue rose 32% year on year to $718 million, net income was $43.7 million for a third consecutive profitable quarter, and active members reached 10.9 million, up 20%. Average revenue per active member was $270, up 10%, while purchase volume grew 19% to $38.5 billion. AI was one of four growth areas alongside Chime Prime, lending and the pending Stride Bank acquisition.

The wider movement of agentic commerce is extending beyond banks to merchants and payment operations. Block and Square’s agentic operations story explores that merchant-side expansion. But account access should depend on confidence in what agents can do and how reliably they do it; Visa’s agentic benchmark underlines why reliability remains central to trust.

What should payments operators do before 24 November?

Payments operators should treat agent access, revocation and liability as design decisions that need to be reflected in customer journeys, contracts and controls.

• Read the agreements you distribute or rely on. Identify what an account provider permits, what it treats as authorised, and what responsibility falls to your business or its customers.

• Set your own agent-access posture. Decide what an agent may access and do, and whether its permissions are limited to information, recommendations or transactions.

• Build revocation and evidence into the product. Customers need a practical way to remove access, confirm that it has ended and retrieve a record of that change.

• Plan for intervention. Consider how an agent’s activity can be paused or stopped if it behaves unexpectedly, and how that event would be investigated.

• Watch the OCC’s controls-first framing. On 5 November 2026, the OCC outlined supervisory pillars for banks delegating to autonomous AI systems, focused on controls and human oversight rather than treating AI as a unique risk category.

• Prepare for the Australian version. The Consumer Data Right and open-banking-style access regimes mean these questions will reach Australia too. Banks and neobanks will need to consider how “open with controls” compares with “open on paper, risk on you” as agents begin to pay bills via PayTo and account-to-account rails.

FAQ

When does Wells Fargo’s AI agent policy take effect? The updated Online Access Agreement takes effect on 24 November 2026.

Can customers use AI agents with their Wells Fargo account? The agreement does not categorically ban them, but says access is at the customer’s own risk and reserves the bank’s right to restrict or block agents.

What is Chime doing with AI agents? Chime is building secure connections for members to use their preferred general-purpose agents alongside Jade, its personal AI finance assistant. It has not announced a launch date.

Are banks required to monitor AI agents? The cited Wells Fargo agreement says the bank does not monitor, screen or review agent activity. The OCC’s November supervisory framework emphasises controls and human oversight for banks’ use of autonomous AI systems.

Key takeaway

Chime is treating AI agent access as a controlled product opportunity; Wells Fargo’s agreement makes customers responsible for managing access and documenting revocation. That difference puts liability, not just connectivity, at the centre of the bank-side debate. For payments businesses, the practical test is whether permissions, evidence and intervention are built into the experience before agents are trusted with money.