Technology•

United Airlines Bars AI Agents From Buying Its Tickets, Even When a Human Approves Every Step

United Airlines changed its Contract of Carriage on November 11 to stop AI agents from booking or managing tickets, citing account security, MileagePlus misuse and federal check-in rules.

Written By :MINRK
United Airlines Bars AI Agents From Buying Its Tickets, Even When a Human Approves Every Step

Photo by CHUTTERSNAP on Unsplash

1. United Airlines AI booking ban takes effect

United Airlines on November 11, 2026 amended its Contract of Carriage to stop AI agents from booking, changing or servicing tickets. The United Airlines AI booking ban covers the carrier's website and app as well as purchases routed through travel agencies and aggregators, and it applies even when a customer supervises or approves each action the software takes.

The change arrives at a moment when software assistants have become capable enough to search for flights, compare itineraries and complete a purchase with little input from the person they serve. Many technology companies have promoted exactly that kind of hands-off booking as one of the clearest everyday uses for so-called agentic AI. One of the largest airlines in the United States has now decided that, at least for its own tickets, the final step must be taken by a person.

United framed the restriction around three concerns: protecting customer accounts and payment details, preventing misuse of loyalty miles, and staying within federal aviation rules that expect passengers themselves to make certain declarations. Chief Executive Scott Kirby has also made clear that commercial considerations play a role, saying the airline does not want a new intermediary standing between it and its customers.

2. What the new contract language says

A Contract of Carriage is the legal agreement every passenger accepts when buying a ticket. It sets out the rules on refunds, baggage, delays, denied boarding and the conduct the airline expects from travelers. Because it governs the relationship between the carrier and the buyer, an amendment to it carries more weight than a line in a help page or a website banner.

The updated terms define a category of automated agents and bar them from carrying out ticket transactions on United's platforms without the airline's prior written permission. The language is deliberately broad. It does not draw a distinction between a fully autonomous assistant and one that pauses to let its user confirm each step. If software is completing the transaction, the rule applies, whether or not a person is watching and clicking approve along the way.

That detail matters because many AI companies have tried to address safety worries by designing agents that seek explicit confirmation before spending money. United's position is that a confirmation dialog is not the same as a human making the purchase. Under the new terms, the human has to be the one who actually completes the booking.

3. Travel agencies and aggregators are covered too

The restriction is not confined to united.com and the United app. It also reaches tickets sold through third-party travel agencies and flight aggregators. In practical terms, that closes off an obvious workaround: a traveler cannot simply point an AI agent at an outside booking channel and have it buy a United seat on their behalf.

Extending the rule this way means United is relying on its distribution partners to respect and enforce the policy. Travel agencies and online booking platforms sell United tickets under agreements with the airline, and the Contract of Carriage applies to the ticket regardless of where it was purchased. How strictly those partners police AI-driven traffic will go a long way toward determining whether the restriction holds in practice.

For travelers, the outcome is straightforward. A chatbot can still be asked to find a cheap flight to Denver, but the purchase itself needs to be made by the traveler, not by the software acting in their place.

4. Why is United restricting AI agents?

United says AI agents create security and compliance risks that it is not prepared to accept. The airline's main concern is that customers who hand an AI assistant their payment details or login credentials expose sensitive information and open the door to unintended purchases, account changes or unauthorized use of MileagePlus miles.

Unintended purchases are the most immediate worry. An agent that misreads an instruction, confuses two dates or picks the wrong fare class can commit a customer to a ticket they never wanted. Airline tickets are often expensive and can come with change fees or fare restrictions, so an error by software can be costly to unwind, and disputes over who authorized a purchase can quickly involve the airline, the customer and a card issuer.

Loyalty balances present a related risk. MileagePlus miles function much like a currency inside United's ecosystem and can be redeemed for flights, upgrades and other rewards. An AI assistant with access to a member's account could spend those miles in ways the member did not intend, and a compromised agent or stolen credentials could expose balances to outright theft. Airlines have long treated loyalty accounts as a target for fraud, and adding another party with account access widens that exposure.

5. Federal check-in rules and the human declaration

The third reason United gives is regulatory. The airline has emphasized that human oversight is needed at check-in so that it stays within rules set by the Federal Aviation Administration and the Department of Transportation on items passengers may not carry.

When travelers check in, they are asked to acknowledge that their baggage does not contain prohibited or dangerous items such as explosives. United's position is that only a person can make that acknowledgment. A traveler knows what is in their own luggage; an AI agent does not. Allowing software to click through that declaration on a passenger's behalf would, in the airline's view, risk running afoul of the federal requirements behind it.

This argument goes beyond commercial preference. Passenger declarations exist so that responsibility for what goes into an aircraft's hold or cabin rests with an identifiable individual. If an automated system were to provide that confirmation, it would be unclear who had actually attested to the contents of a bag, and regulators would be unlikely to treat a chatbot's checkbox as equivalent to a passenger's own statement.

6. A preemptive move rather than a reaction

United has not yet seen a meaningful volume of bookings made by AI agents. The policy is therefore an attempt to set the rules before the behavior becomes common, rather than a response to a wave of problems already on the airline's books.

Acting early gives the carrier a cleaner position. Writing restrictions into the Contract of Carriage now means that if AI booking tools become widespread over the coming months, United will already have the contractual basis to refuse or reverse transactions it considers unauthorized. Waiting until agent traffic was substantial would have made a change far more disruptive for customers who had come to rely on it.

The timing also reflects how quickly the technology is advancing. Assistants that can navigate websites, fill in forms and complete checkouts have moved from demonstrations to consumer products in a short span, and companies that handle payments and personal data are being forced to decide how far they are willing to let that software operate on their platforms.

7. Scott Kirby and the middleman question

Security and compliance are not the whole story. Kirby has been candid that United also wants to protect its direct connection with the people who fly with it. "To be honest, I don't want to create another middleman," he said.

That comment speaks to a long-running tension in the airline business. Carriers have spent years trying to steer customers toward their own websites and apps, where they control how fares are displayed, can promote seat upgrades and extra services, and collect data that feeds their loyalty programs. Every intermediary that sits between an airline and a passenger, from traditional travel agencies to online booking sites, takes a share of that relationship and often a share of the revenue.

AI agents threaten to become a new and potentially powerful layer in that chain. If travelers began delegating all of their flight shopping to an assistant, the assistant rather than the airline would decide which options to present and how. An agent optimizing purely for price could also bypass the add-ons and branded experiences airlines use to differentiate themselves. By keeping the purchase in human hands, United preserves the moment at which it can speak directly to its customer.

8. Part of a wider pullback from AI shopping agents

United is not alone in drawing a line. Amazon and some large banks have also restricted AI agents from operating on their platforms, pointing to similar concerns about security and control over customer accounts.

The pattern suggests that the companies holding customers' money, credentials and purchase histories are becoming more cautious just as AI developers push to make their agents more autonomous. Businesses that run large consumer platforms face a choice between accommodating a new channel that could bring in sales and guarding against the fraud, disputes and loss of customer contact that the channel could also bring. For now, a growing number are choosing caution.

For people who build or use AI tools, the restrictions are a reminder that an agent's capabilities are only part of the picture. Whether an agent can complete a task also depends on whether the business on the other side is willing to let it, and contractual terms like United's give companies a formal way to say no.

9. Enforcement challenges and what comes next

Writing a rule is easier than enforcing it. A capable AI agent operating through a normal web browser can be difficult to tell apart from a human customer, and the restriction on purchases made through travel agencies depends on those partners detecting and refusing agent-driven transactions. How United identifies prohibited activity, and how consistently its partners apply the policy, will determine how much the change alters behavior.

What United has established is a clear principle: wherever a ticket purchase creates a legal or financial obligation, or wherever federal rules call for a passenger's own declaration, a human must be the one to act. Travelers remain free to use AI to think through their options before booking, but the commitment itself now belongs to the person who will be sitting in the seat.

The broader question is whether other carriers follow United's lead or take a more permissive approach in the hope of attracting AI-driven bookings. Either way, the new contract language marks one of the first instances of a major US airline formally deciding how much authority it will grant to software acting on a passenger's behalf.

Related Articles

NEWSLETTERS

Don't miss another story.

Subscribe to the MINRK Newsletter today.

Tick the newsletters you want. No selection? You'll get MINRK Headlines.

By signing up, you will receive emails about MINRK products and you agree to our terms of use and privacy policy.

MINRK Daybook

Market analysis for crypto traders and investors.

EVERY WEEKDAY

MINRK for Advisors

Defining crypto, digital assets and the future of finance for financial advisors.

EVERY THURSDAY

MINRK Protocol

Exploring the tech behind crypto one block at a time.

WEEKLY

MINRK Long & Short

A must read for institutions. Insights, news and analysis delivered weekly.

EVERY WEDNESDAY

MINRK Headlines

The biggest crypto news and ideas of the day.

EVERY WEEKDAY

MINRK Policy

Examining the intersection of cryptocurrency and government.

WEEKLY

MINRK Research

Join thousands of readers who rely on MINRK for data-driven insights on the latest digital asset trends.

MONTHLY