Websites Are Locking Out AI Shopping Agents, and Banks Are Following
Meta's Muse AI agent ran into website verification checks at Walmart, one of its retail partners, while Amazon deliberately blocked the tool from shopping on its site. The episodes, reported by Engadget on 2 November 2026, put agent access control at the center of a dispute that is spreading from retail into financial services.
Walmart said its verification checks were not intended to stop the Muse assistant. The checks nonetheless presented Muse with a barrier on a site where Meta has a partner relationship. Amazon took the opposite approach and blocked Muse from shopping on its platform.
Amazon's position extends beyond Muse. The company contends that Perplexity's Comet AI browser violates its access rules and puts customer data at risk. Perplexity maintains that users have authorized its Comet agent to act on their behalf. A federal appeals court vacated a preliminary injunction, a court order that temporarily pauses a party's conduct while a case proceeds, against Perplexity in August 2026, but the underlying Amazon-Perplexity disagreement remains unresolved. Amazon also offers its own Buy for Me feature, which can purchase products from third-party brands' websites. The company is therefore both restricting outside agents and operating one of its own.
Other merchants are drawing lines in their terms. eBay's user agreement prohibits unauthorized buy-for-me agents unless eBay grants permission. Delta stated it has no integration that allows third-party agents to book flights through its website. Delta did not confirm whether it deliberately blocks specific AI agents.
The banking sector has been more explicit about the technical concern. During JPMorgan Chase's October earnings call, CFO Jeremy Barnum said some AI agents use shared login credentials and screen scraping (reading the text and layout a browser displays, much as a person would) to transfer funds. He described those methods as "not secure." JPMorgan Chase blocks them where possible. Barnum also warned that customers who hand over credentials could be responsible if an AI agent takes an unwanted action. According to Reuters, published on 22 October 2026, major U.S. banks have been blocking third-party AI agents from operating inside customer accounts and making transactions. Earlier, on 22 September 2026, Reuters reported that NatWest warned that using AI agents for online shopping could increase the risk of scams, fraud and data-privacy breaches.
The technical core is the same in every case. An agent that logs in with a user's password, or drives a browser by scraping the rendered page, is indistinguishable from the user at the transport layer (the network level where requests are sent). It inherits whatever authority the session carries. Merchants and banks see no reliable way to separate authorized delegation from abuse, so they default to blocking. The verification checks Muse met at Walmart and the blocks described at Amazon are the visible edge of that default. The banking statements add a liability dimension, because the credential holder is the party Barnum said could be responsible.
In my view, the stakes of this dispute depend on whether a sanctioned path emerges. Blanket blocking protects platforms and accounts, but it also pushes agent builders toward scraping and credential sharing, the very methods the banks call insecure. The eBay model, in which the platform grants permission to specific agents, points toward a different outcome, one where delegation is explicit, scoped and revocable. Whether merchants, banks and agent developers converge on a standard for that kind of permissioning is the question the current disputes have not yet answered.
