United Airlines says it is taking steps to restrict unauthorized third-party personal AI assistants from interacting with its travel services. The Nov. 11 announcement describes controls inside United’s digital platforms and through travel agencies and aggregators, alongside updates to booking, ticketing and distribution policies. It is a policy direction and work in progress: United has not published a rollout date, a technical list of blocked tools or a completed blanket ban.
What is United changing?
The airline distinguishes its own app, where customers can manage bookings and seats, from personal agents that act through a customer’s account or third-party travel channels. United says it intends to prohibit use by unauthorized agents. The announcement does not name particular assistants, describe an API change or explain how it will identify each interaction. Its reference to indirect restrictions through agencies and aggregators signals that enforcement may extend beyond the airline’s website and app.
That leaves practical questions unanswered for travelers and developers. United has not specified how it will define authorization, whether a customer can approve an agent, or whether any supported integrations will remain available. It also says it plans to update its Digital Trust Center with more detail and help for customers if an agent takes an unintended action. Until that guidance appears, readers should not assume a particular booking flow is either supported or already blocked.
Why does the airline say it is doing this?
United points to several risks. Sharing credentials with an assistant could expose account information. An automated action could make a purchase, alter a reservation or use MileagePlus miles without the customer intending that outcome. Those are familiar authorization and recovery problems for software that can click through a logged-in interface: a request can be misunderstood, duplicated or completed at the wrong time.
The company also raises a narrower regulatory example: check-in requires a human acknowledgment that baggage does not contain prohibited substances such as explosives. United says automated actions may have the potential to violate FAA and Department of Transportation rules. Its statement does not cite a specific agency determination that every AI-assisted action is unlawful; it presents regulation as one reason for the proposed restrictions.
What should customers and builders do now?
For travelers, the immediate advice is conservative: use United’s own service for bookings and account changes, avoid giving a third-party assistant account credentials, and review itineraries and mileage activity directly. These steps reduce the chance of an agent acting beyond the person’s intent while the airline clarifies its policy. They do not imply that United has already disabled every automated workflow.
For makers, the announcement illustrates why an agent needs an explicit permission boundary around purchases, identity checks and irreversible account changes. A reliable product should show the proposed action, identify the account it will affect, request human approval at meaningful checkpoints and provide a usable audit trail and recovery path. Access controls should also match the service’s own rules; a customer’s password alone does not grant a tool permission from the airline.
United’s post is an initial statement rather than a full technical policy. The open questions are authorization, enforcement and supported alternatives, and the airline says more customer guidance is coming. VIBPS has not tested a blocked booking flow. We will need the airline’s forthcoming trust-center details to assess what changes in practice.