--- title: "Why Companies Are Blocking AI Agents: What United Airlines’ Ban Means for Your Business" source: https://iaastha.com/insights/blog/why-companies-are-blocking-ai-agents-what-united-airlines-ban-means-for-your-business/ type: Post date_published: 2026-11-14 date_modified: 2026-11-14 author: Sarah description: "United Airlines now restricts unauthorized third-party AI agents from booking flights, citing shared logins, unintended purchases and federal rules. The wider lesson is simple: businesses will let AI…" publisher: iAastha Research & Consulting --- # Why Companies Are Blocking AI Agents: What United Airlines’ Ban Means for Your Business United Airlines now restricts unauthorized third-party AI agents from booking flights, citing shared logins, unintended purchases and federal rules. The wider lesson is simple: businesses will let AI agents in only when they can verify who the agent is, limit what it can do, and name who is accountable when it makes a mistake. An airline just told [AI agents](https://iaastha.com/) they cannot book a seat. At first glance that sounds like a travel story, but it is really the first clear answer to a question every business will face soon: who is allowed to act on a customer’s behalf, and who is responsible when software does it? In this guide we break down what United announced, why companies are blocking AI agents, how other airlines and banks are responding, and what it means if you build or use autonomous AI agents for business. ## **What did United Airlines actually ban?** In November 2026, United said it will restrict access for AI agents that try to book flights and manage reservations for customers. The limits apply directly on its own website and app, and indirectly through travel agencies and aggregators. The target is what United calls unauthorized, third-party personal AI agents or assistants. The airline said this work includes updates to its booking, ticketing and distribution policies. The rule is also written into United’s revised Contract of Carriage, which defines an AI agent broadly as software that takes autonomous or semi-autonomous action, even if a passenger supervises it. In plain terms, a traveler can ask an assistant to help, but United does not have to let that assistant complete the transaction. It helps to know what the ban does not cover. Agents can still show United flights and prices, and a customer who researches with an AI tool and then buys directly is not breaking the rules. United has also said it is not seeing many agent bookings today, so the policy is as much about where travel is heading as where it is now. ## **Why are companies blocking AI agents?** United gave several reasons, and each one applies far beyond aviation. Together they explain most of the current wave of AI agent restrictions. ### **1. Shared credentials create AI agent security risks** To book a ticket, an agent needs a customer’s login, personal details and payment information. Handing those to a third-party tool multiplies the places where they can leak. United’s own warning is blunt: *“sharing account credentials with personal, third-party, AI assistants can expose sensitive information”* ### **2. Agents can take actions nobody intended** United warned that agent actions may lead to unintended purchases, account changes or the use of MileagePlus miles. A human who mistypes a date notices the error. An agent that misreads an instruction can repeat it across many accounts before anyone looks. ### **3. Some legal steps need a human** Federal aviation rules require a person to confirm that their bags do not contain banned items at check-in. An agent cannot honestly make that declaration. This is the clearest example of a task where the law assumes a human, and it shows how many business processes quietly rely on that assumption. ### **4. Control of the customer relationship** Aviation analyst Henry Harteveldt of Atmosphere Research Group said the policy suggests United wants to keep control of the booking process. That is a commercial reason, not a security one, and it matters. If an agent becomes the front door to every purchase, the business behind it risks becoming a supplier to someone else’s interface. ## **Can AI agents book flights right now?** It depends on the airline, which is exactly why this story is worth watching. Delta has signaled that assistants can complete bookings once they are authenticated, with extra safeguards for sensitive trips such as unaccompanied minors. American has also indicated that agents can book in certain circumstances. United has chosen the strictest position. This split creates a practical question for anyone building a booking or shopping agent: if only some providers let the agent close the sale, the agent will naturally favor those providers. Access rules will shape which businesses agents recommend, which makes them a competitive lever as well as a safety measure. ## **It is not just airlines** United is the latest large company to push back. Banks are making the same argument. Bank of America recently urged customers not to share their banking passwords with AI assistants. In retail, Amazon has blocked Meta’s Muse agent while Walmart chose to partner with Meta instead, so even competitors in the same industry disagree on the right approach. Real-world usage is still small. One weekly industry roundup reported that during Uber’s Halloween week, agents placed about 500 Uber Eats orders and four ride bookings out of roughly 320 million total trips. The volume is tiny today, which is exactly why this is an inexpensive moment for companies to set rules before agent traffic grows. The common thread is that businesses are struggling to tell a real customer from a piece of software acting for that customer. A contract can ban agents, but a contract is not a technical control. The lasting fix will be technical. ## **AI agent identity and permissions: the real problem** Strip away the headlines and the issue comes down to three questions that every system must be able to answer about an agent. 1. **Identity:** can the business tell who or what is acting, and who sent it? 2. **Permission:** does the agent have only the access the task needs, for a limited time? 3. **Accountability:** when something goes wrong, who owns the mistake, and can anyone trace it? Work is already underway. Visa, for example, is developing ways for merchants to recognize approved agents and connect their actions with customer and payment information. United itself says it wants standardized, secure integrations with authorization and accountability protections before it lets agents interact directly with accounts. The direction is clear: agents will be welcome when they arrive with credentials of their own, not borrowed passwords. ## **Try it: the agent trust checker** Use this quick checker to test whether an AI agent you use or plan to build would pass the three tests above. Tick everything that is true today. The agent has its own identity and does not log in with a person’s shared password. Its access is limited to the specific task, with no broader account rights. Every action is logged so it can be traced and reviewed later. A human must approve payments, deletions and other irreversible steps. A named person or team owns the agent’s mistakes. ## **What businesses should do about autonomous AI agents** If you run a business, treat this as a prompt to decide your agent policy before customers force the question. Here is a practical starting point. - **Decide what agents may do.** Allowing research and price display is very different from allowing purchases and account changes. Write the line down. - **Never share personal passwords with a tool.** Use scoped tokens or delegated access so that one leak does not expose an entire account. - **Keep a human on irreversible steps.** Payments, deletions and legal declarations should need a person’s approval. - **Log everything.** If you cannot trace an agent’s action, you cannot assign responsibility for it. - **Plan an official path.** Blocking is a short-term answer. Businesses that offer a secure, authenticated way for agents to connect will be invited into the next wave of AI-driven commerce. If you are building AI agents for business, the same advice applies in reverse: design for identity, permission and accountability from day one, and you will find far fewer doors closed to you. ## **Frequently asked questions** **Why are companies blocking AI agents?** They cite shared login credentials, unintended purchases or account changes, rules a machine cannot satisfy, and a wish to keep control of the customer relationship. **Can AI agents book flights?** It depends on the airline. United restricts unauthorized third-party agents, while Delta and American have signaled that agents can book in certain circumstances. Agents can still research flights and prices on United, and customers can then buy directly. **What is AI agent identity?** It is the ability of a business to verify that a software agent is approved and to link its actions to a real customer, instead of relying on a borrowed password. **Are AI agents safe to use with my accounts?** They carry real risk when they hold your credentials or can act without approval. Limit their access, avoid sharing passwords, and require human approval for payments. --- Cite as: "Why Companies Are Blocking AI Agents: What United Airlines’ Ban Means for Your Business" — iAastha Research & Consulting, https://iaastha.com/insights/blog/why-companies-are-blocking-ai-agents-what-united-airlines-ban-means-for-your-business/ Site index for AI: https://iaastha.com/llms.txt