On 11 November 2026, United Airlines updated its Contract of Carriage, adding Rule 6.J.6, which explicitly prohibits the use of third-party AI agents to book, modify, or service tickets. The airline explained the move by citing the risks of data leaks, unintended purchases, and violations of FAA and DOT rules requiring a passenger to personally confirm the absence of prohibited items in their baggage.
The prohibition mechanism affects not only direct access to united.com and the app but also indirect channels — through travel agencies and aggregators. AI agents such as Meta Muse or OpenAI Dots can still search for flights and compare prices, but a human must complete the transaction. This distinguishes United's approach from that of Delta, American, and JetBlue, which leave the risk with the customer provided authentication is performed using their credentials.
Technically, agentic AI requires the transfer of credentials — login, password, payment information, and access to MileagePlus. Unlike simple chatbots, such agents perform multi-step actions: searching, comparing, selecting seats, paying, and even check-in. United emphasizes that handing credentials to third-party systems creates a point of failure where a leak or an agent's error leads to unauthorized mileage deductions or booking changes.
Methodologically, the ban is based on regulatory requirements, not merely on internal policies. The FAA and DOT stipulate that a passenger personally confirms baggage safety; AI cannot provide such a legally significant confirmation. This creates a gap between the capabilities of modern agentic systems and the real requirements of high-risk industries, where authentication and compliance are not reducible to API calls.
Compared with other sectors — banks, retailers such as Walmart and Amazon — United has become the first major airline to explicitly write the prohibition into its contract of carriage. CEO Scott Kirby stated plainly that he does not want to create "yet another intermediary." This is a strategic choice: to retain direct control over the customer experience and loyalty data instead of integrating with external agents.
The implications for the industry are obvious: developers of agentic AI will be forced either to obtain official partnerships with airlines or to limit themselves to a research phase without a final transaction. For passengers, this means that the convenience of delegating booking is for now limited to search rather than purchase.
The question remains open as to how quickly standards for secure delegation will emerge — for example, through OAuth-like protocols with limited permissions or cryptographic verification of an agent's actions. For now, United is demonstrating that in heavily regulated spheres, security and legal liability still outweigh the speed of automation.
United's ban shows that the maturity of agentic AI is measured not only by its ability to perform tasks but also by its ability to meet industry requirements for authentication and compliance without creating new risks.
