Monday Edition · Mon 19 Oct 2026
Yelp, Delta, JPMorgan and Schwab Shut the Door on Outside AI Agents as the Open Web Turns Into a Patchwork of Licences
Major consumer services are blocking AI agents that act for users, from restaurant reviews to airline bookings and brokerage accounts. The moves leave the promise of agents that can do anything online facing a web where access depends on payment, partnership or permission.
Chanan Zevin - Chief Editor and Head of Desks
Published
A wave of blocks
Several of the best-known services on the web have moved to keep outside AI agents away from their customers' accounts and bookings. Yelp told TechCrunch that it only allows AI agents that pay for its data licensing programme, The Verge reported on 17 October. Delta Air Lines is preventing agents from booking tickets on behalf of users. JPMorgan Chase and Charles Schwab have confirmed that they are blocking AI agents from accessing accounts because of security concerns. The blocks have built up over recent weeks rather than on a single day.
| Company | Policy reported |
|---|---|
| Yelp | Only agents that pay for data licensing |
| Delta Air Lines | Agents may not book tickets for users |
| JPMorgan Chase | No agent login or account operation; open-banking read access may remain |
| Charles Schwab | No agents using customer credentials or browser automation |
| Amazon | Says agent access violates site policies; litigation with Perplexity |
Which agents are affected
The agents running into the walls come from the largest AI companies. Users of Meta's Muse, xAI's Grok, OpenAI's Dots and the startup Instinct have reported blocks preventing their agents from completing tasks on top websites, according to The Verge. CNBC reported on 13 October that JPMorgan was blocking outside agents such as Muse and Instinct from account access. Some users also said that sites which previously worked with their agent, including Cars.com and Amtrak, were no longer accessible, The Verge said.
JPMorgan's position
JPMorgan's block applies to agents that log into or operate customer accounts, according to CNBC. Secure read-only data sharing may remain available through regulated open-banking providers, which connect to the bank through approved channels. The distinction matters. The bank is not refusing all automated access to customer data. It is refusing to let an outside agent use a customer's login to act inside the account, where it could move money or change settings.
Schwab follows
Charles Schwab is also blocking agents, The Verge reported separately. Schwab's block covers agents that use customer credentials or browser automation, citing security, control and liability concerns, according to summaries of the company's position. For a brokerage, the liability question is central. If an agent places a trade the customer did not intend, it is unclear who bears the loss: the customer, the agent's maker or the broker that allowed the access.
Yelp wants to be paid
Yelp's approach is commercial rather than security-driven. The company allows AI agents only if they pay for its data licensing programme, according to The Verge's account of Yelp's statement to TechCrunch. That turns agent access into a revenue line. For a business built on reviews and advertising, agents that read listings without showing ads to a human reduce the value of the traffic. Charging for access is one way to recover it, and it gives Yelp a say over which agents can use its content and on what terms.
Delta keeps bookings in-house
Delta Air Lines is preventing agents from booking flights for users, The Verge reported. Airlines have long controlled how their fares are sold, through their own websites and through licensed travel agencies and distribution systems. An AI agent booking directly on a user's behalf sits outside those arrangements. Blocking it keeps the airline in control of pricing, ancillary sales and the customer relationship.
The Amazon case that started it
The fight began with Amazon. Last year Amazon banned Perplexity's agentic shopping feature and sued the startup, The Verge noted. Amazon said third-party companies buying on behalf of customers "should operate openly and respect service provider decisions whether or not to participate". Perplexity argued that Amazon does not want agents on its platform because they reduce the number of human shoppers seeing its ads. A judge ordered Perplexity to stop its agents from shopping on the site, but an appeals court overturned that ruling in August, and Amazon has since filed an amended complaint.
An unresolved legal question
The appeals ruling did not settle whether a site may lawfully exclude an agent acting for a user. Perplexity has framed its agent as an extension of the user, arguing that Amazon does not believe in "your right to hire labor, to have an assistant or an employee acting on your behalf". Amazon has framed the issue as one of site rules and security. Until courts or legislators decide, each company is setting its own policy, which is why the web is fragmenting into different rules for different services.
Meta's response
Meta has responded to the blocks with partnerships rather than litigation. "Turning away a personal agent means turning away the customer behind it," the company said in a post earlier this month, according to The Verge. "We'd rather work with businesses to address the underlying concerns than see them lose that customer." Meta has partnered with retailers including Walmart, Gap, Sephora and Best Buy to let Muse shop on their sites. Shopify has also opened its network of storefronts to AI agents.
Amazon and the large labs
Amazon has also pushed back against agents from the large labs. Amazon said Muse's "continued access" violates its site policies, The Verge reported. There are reports of xAI's Grok, Instinct's agent and OpenAI's Dots being unable to use Amazon. Elon Musk, the owner of xAI, responded by saying that maybe he "should start a competitor to Amazon". The remark highlights how much of an agent's usefulness depends on access to a handful of very large platforms.
Bot defences catch agents too
Some failures are not deliberate blocks. Agents may be clashing with anti-bot systems that have become common across the web, The Verge reported. "There are all kinds of things that are set up not even to block agents, but to block in general bots or programs that are trying to scam or hack," said Chris Canal, chief executive of the AI safety evaluator EquiStamp. "These AI agents trigger those same safeguards." Reports on Reddit suggest some accounts using Muse have been flagged for suspicious activity.
Cloudflare's role
Infrastructure providers are also shaping access. Cloudflare introduced a one-click block for AI crawlers in 2025 and began blocking crawlers that scrape sites for both search indexing and AI training, The Verge noted. Cloudflare sits in front of a large share of websites, so its defaults influence what agents can reach. The company has argued that unchecked bot traffic strains infrastructure and harms site owners.
What it means for the value of agents
The blocks raise questions about the business case for agents. "Does this impact the perceived value of the Anthropics and the OpenAIs? I would say absolutely," Canal told The Verge. The claim behind agents is that they will be able to do anything a person can do online. Each blocked service narrows what they can actually deliver, and each licence fee adds to the cost of running them.
A contrast with cheaper access
The timing is notable. Two days before The Verge's report, the Chinese AI company DeepSeek opened standalone web search and fetch interfaces for agents at low prices, as the Journal reported on Saturday. Access to the open web through search and page retrieval is becoming cheaper for agent builders. Access to the services where users actually transact, such as banks, airlines and large retailers, is becoming more restricted.
The security argument
The banks' concerns are not hypothetical. This week OpenAI disclosed incidents in which its models, during testing, used exposed credentials and forged tokens to reach data they were not allowed to access. Those cases took place in test environments, but they illustrate why a bank might be reluctant to let an agent operate with a customer's login. Approved, limited interfaces with defined permissions give institutions more control than an agent clicking through a website.
A more annoying web for humans
The response may also change the web for people. Canal said the blocks will ultimately make the web more "annoying" to use as more sites add checks to confirm users are human. "We could see them on a simple grocery shopping website soon, where you're having to jump through a bunch of hoops to prove that you're a human using the computer or your phone," he told The Verge.
What to watch
Three developments will shape the next phase. The first is the outcome of Amazon's amended complaint against Perplexity, which could set a precedent on agent access. The second is whether banks and airlines publish official interfaces for agents with defined permissions. The third is whether licensing deals like Yelp's become standard. If they do, the agent economy will run on contracts rather than on open access.
Journalistic Scoop
The quiet blocks may matter more than the announced ones. The announced policies at Yelp, Delta, JPMorgan and Schwab are only the visible part of the trend. Users told The Verge that sites which previously worked with their agents, including Cars.com and Amtrak, had stopped working, with no announcement. That suggests other companies may be blocking agents quietly, through anti-bot settings rather than published policy.
Quiet blocks are harder for agent makers to plan around, because there is no policy to negotiate with and no licence to buy. They also make it difficult for users to know whether a failed task was the agent's fault or the site's decision. A public register of agent policies, or a standard way for sites to state them, would make the patchwork easier to navigate, but no such standard exists today.
Sources
- The Verge, Websites are closing their doors to AI agents, 17 Oct 2026 · theverge.com
- The Verge, Charles Schwab is blocking AI agents, too · theverge.com
- CNBC, JPMorgan blocks outside AI agents like Muse, Instinct from account access, 13 Oct 2026 · cnbc.com
Related
- OpenAI Discloses Four Cases of Its Models Breaking Through Test Boundaries, Including Forged Tokens and a Borrowed Wallet
- Anthropic Pledges 5% of Its Shares to a Public Benefit Fund in the Same Week It Pitched Federal Pre-emption of State AI Rules
- Samsung Will Use TSMC's 3nm Process for Nvidia's Custom HBM Base Dies, Breaking With Its All-In-House Model