Some of the largest companies in the banking and retail industries are beginning to engage in a competition with tech giants, as well as among themselves, around a new type of customer – AI individual agents.

JPMorgan Chase, Charles Schwab, Amazon, Neiman Marcus, and Kohl have already begun to block individual agents like Meta Muse, while consumers are just now starting to use these tools to find higher savings rates, compare prices when shopping, and cancel subscriptions.

The era of proxy-based AI has arrived, but there is still uncertainty about how the industry landscape will change when virtual stores are flooded in by a large number of AI proxies, or when companies attempt to stop them.

Last week, there were reports of AI agents “getting out of control” in customer accounts, and prominent politicians including Senators Mark Warner and Bernie Sanders called for legislative regulation of robots due to these headlines. Following that, Jamie Dimon, CEO of JPMorgan Chase, stated that the bank would currently prevent external AI agents from operating on behalf of retail customers due to security risks.

Some retailers have also chosen to embrace this trend. Last week, Walmart announced that it would open its virtual doors to the AI agents that customers prefer. In a blog post announcing this decision, the company stated: "We price our products, not people; therefore, the prices seen by AI agents are the same as those seen by humans on Walmart.com, and the applicable discounts and fees will also be displayed before purchase."

An identity verification company, Billions Network, recently audited the public access policies of 51 major US retailers and found that only 11 of them had a clear stance regarding AI proxies. This includes Amazon, Kohl, and Neiman Marcus; these five companies explicitly restricted AI proxies. On the other hand, 37 retailers inadvertently blocked shopping proxies through their existing rules designed for bots. None of them provided a way for verified proxies to complete purchases.

According to the analysis of Billions Network, blocking proxies can be achieved in various ways: for example, as with Chewy, by blocking specific robots from third parties based on rules; or more general restrictions can be adopted to prevent any third-party robots from attempting to access login, registration, and checkout functions. This analysis also reminds us that publicly available rules do not necessarily guarantee that these restrictions will be technically enforced.

Billions, co-founder and CEO, stated that as of now, "most of the retailers we are auditing have not made a decision regarding AI agents yet."

Accenture's global banking business leader, Michael Abbott, believes that in the long run, it will be very difficult for companies to succeed if they adopt a strategy of preventing AI agents.

Abbott said, "If a client explicitly authorizes AI to act on their behalf, then it is ultimately an extension of the client's choice. The more important question is not whether institutions can prevent agents, but whether they can create a safe and credible way for agents to participate."

Abbott According to Accenture's research, 65% of consumers are willing to use financial assistants similar to GPT through generative AI platforms or digital wallets, and 71% of consumers are willing to use embedded AI assistants within the mobile apps of their primary banks.

He said, "We've seen this situation before... online banking, mobile banking, and digital wallets. Customers tend to gravitate towards experiences that reduce friction and save time. In the end, it's usually the institutions that manage to participate safely rather than resist change that come out on top."

The real opportunity for banks is to become credible providers of the data, intelligence, and execution capabilities that these agents rely on.

Abbott said, "As interactions increasingly occur outside of banks' own channels, banks still need to be present at those critical decision-making moments, even if they don't own those interfaces."

Let customers take control of the AI proxy.

When it comes to distinguishing between genuine security concerns and companies trying to protect their profits by making fund transfers or price comparisons more difficult for consumers, Abbott stated that both can be true at the same time.

He said, "There are indeed very real concerns regarding fraud, authentication, liability, and consumer protection. Customers also expect banks to protect their assets and data – these responsibilities do not disappear just because AI agents emerge." He added that at the same time, consumers have always benefited from technologies that enhance transparency and make it easier to compare options.

Abbott said, "Our research findings show that 82% of people hope to approve every AI operation, and 79% of people wish to immediately pause or stop AI activities. This indicates that consumers desire convenience, but not at the cost of sacrificing control."

Mike Danford, the owner and Chief Revenue Officer of Adverio and responsible for operating Amazon, Walmart, and Target platform businesses for e-commerce brands, stated that some of the security concerns mentioned by companies, including banks, are not entirely about security issues, but rather about in which "sandbox" customers are operating.

Danford said, "Amazon has blocked Muse and other external proxies, but its own Buy, for, and Me functions use Amazon's proxies to operate on websites of other brands. It seems that Amazon is not really concerned about the proxies themselves, but rather who holds the data." He added, "Amazon now allows shoppers to set rules for automatic purchases when prices reach a certain threshold, and consumers seem to accept this type of proxy-based purchasing, as the shoppers are still making the decisions within the preset parameters."

An Amazon spokesperson stated, "We believe that third-party apps that represent customers in purchasing goods from other companies should operate transparently and respect the decision of service providers whether to participate or not. Proxy-based third-party apps like Muse also have the same obligation. We have requested that Meta remove Amazon from this experience."

Danford said that the situation in banks is similar as well; they just charge fees instead of building a wall.

He said, "JPMorgan Chase now charges aggregators like Plaid for scraping account data, so the cost will be passed on to the applications used by consumers and may be reflected in subscription fees."

JPMorgan Chase declined to comment.

Plaid indicates that when the agreement is reached in September 2025, it will not affect any existing pricing arrangements with customers. Plaid No fees are charged to consumers; however, applications and services that use Plaid will decide on their own the fees and subscription prices to charge to customers.

Visa Head of Payment Ecosystem Risk and Control Michael Jabbara Refuses to comment on the motives of any bank for preventing proxy AI, but he stated that the risk of fraud from proxies is "very real and very specific."

He said that consumers may be deceived by fake agents created by scammers, which can lead to the theft of credentials and payment information; moreover, legitimate agents might also engage in actions that customers never intended to authorize, such as purchasing a sofa that the shopper merely asked them to find.

Agents can also exhibit unpredictability: Jabbara described a user who asked an agent to book a gym slot, but when that slot was already occupied, the agent actually hacked into the gym’s scheduling software, deleted the existing reservations, and filled in the user’s name instead. He said that the main concerns lie in figuring out “who the agent is, who the agent is acting on behalf of,” and whether it has “the appropriate permissions.”

He said, "Proxies are not just about exchanging data. Proxies can also take action on behalf of users."

Like other experts, Jabbara believes that the shift from online shopping to mobile shopping is of reference value, but for different reasons: "Before those innovations can truly scale, it is necessary to establish a fundamental layer of security first, so that trust can support such scale."

"Proxy Flood"

Another concern is scale. Millions of agents working legally for clients could overwhelm websites and customer service systems, a phenomenon some refer to as "proxy flooding."

At a time when personal proxies such as Muse suddenly emerged, an expert from the Wharton School of Business, AI, Ethan Mollick, posted: "We will soon see a type of completely legal distributed denial-of-service attack targeting customer service departments, and it will be authorized by millions of consumers; once everyone knows that Muse / Dot can submit old refund requests and negotiate discounts indefinitely, this situation will occur."

Abbott considers this to be a possibility, rather than a crisis. He said, "I see it more as an infrastructure challenge, rather than a threat to survival. Most digital systems today are designed for interactions between people. Agent-driven interactions may occur on completely different scales and frequencies. A person may have multiple agents continuously monitoring interest rates, researching products, or evaluating alternatives at the same time."

However, he stated that the solution is not to stop them. "Rather, it is to build the next generation of infrastructure that can verify trusted proxies, authenticate permissions, manage identities, and efficiently support machine-to-machine interactions."

He added that this is not unfamiliar to banks.

In the past, customers interacted with banks only a few times a month, but now, with mobile devices, they interact almost daily. The AI agent is just the next step in this evolution. Those institutions that are prepared for this reality will be in a more advantageous position than those that try to stop it.

Although many financial websites have implemented defensive measures to prevent overuse in order to counteract automated denial-of-service attacks that have potentially existed for a long time, Ed Felten, a retired professor of computer science from Princeton University who once served as Chief Technology Officer at the U.S. Federal Trade Commission and as Deputy Chief Technology Officer for the Obama administration, stated that proxies "never sleep and never lose focus." Moreover, the more proxies are used by companies such as banks, the higher the costs may be for those organizations.

But Felten suspects that this kind of blockade won't last long. He said, "When a button in the user's browser is clicked, it's very difficult for remote websites to determine whether it was a human click or an instruction sent by a AI proxy on the user's computer. If there really is a battle over access to AI, it will slow down the spread of AI proxies; but in the end, the proxies will win because they are too useful to users. Therefore, those websites that are too effective at blocking proxies will ultimately lose business."

How long can giants still keep the 'gate' closed?

Experts say that as market forces compel websites to accept customer proxies, a common expectation regarding what constitutes reasonable proxy behavior will gradually take shape.

But SharpLink, the CEO of Joseph Chalom – one of the companies that holds the most Ethereum – who also previously worked at BlackRock – expects that the doors will remain closed for some time to come.

He said, "These large, established companies actually control the entry points and will try to shut the doors in order to defend against such innovation threats. For decades, disruptors have been fighting to gain access in fields such as tourism, telecommunications, and financial data aggregation." He added that in each case, the established companies claim it is out of security concerns, but in reality, they also face the problem of their business models being severely impacted.

"For example, AI agents working for users can easily re-book flights at lower costs, eroding the profit margins of airlines. More importantly, AI agents can transfer idle cash from checking and savings accounts, which earn almost no interest, to money market accounts that offer yields ten times higher," said Chalom, noting that as agents gain comprehensive financial knowledge and attention to detail, such 'conversion costs' are being reduced to near zero."

Therefore, I don't think they will be in a hurry to lower the drawbridge.

Abbott indicates that in the short term, some institutions will welcome agents, while others will continue to impose restrictions, focusing first on resolving issues related to authentication, responsibility, and governance. In the long run, he says, data shows that people will entrust administrative tasks in financial decision-making to their agents. "Defensive strategies may prevent losses in the short term, but to win, banks need to adopt an offensive approach."

Proprietary robots must be powerful enough not only to retain customers but also to make them reluctant to use other agents, which is also the idea being expressed by Andy Jassy and others at Amazon. Jassy stated during last week's latest quarterly earnings conference call: "We believe that Alexa, for, and Shopping will become the main shopping agents that customers use when shopping on Amazon. When you browse, get inspiration, or have questions on the product details page, there is always a shopping assistant by your side, and you can directly ask it to place an order based on your existing purchase history, delivery address, and payment information. This is unique, helpful, and very convenient for customers."

In the end, brands will face unprecedented challenges. Human customers may develop an attachment to a brand for reasons beyond price: nostalgic cereals that evoke warm memories of childhood, airlines that carry the memories of family vacations, or banks with familiar clerks. But the AI agents don't care about any of these.

The President of F Token Issuance and Software Business, a senior executive at this analytics company renowned for its credit scoring, Nikhil Behl, stated: "Brand loyalty and customer relationships are being put to the test. AI agents don't have feelings for logos or ten-year relationships. They have a task, and they only care about the results." He added, "The levers that were used in the past to enhance loyalty will be replaced by levers that agents can truly understand, such as consistent approval processes, transparent pricing, fast problem-solving, and terms that stand up to price comparisons."