Hospitality · 17 November 2026
United Airlines Bans Third-Party AI Agents From Flight Booking Accounts
United Airlines has barred customers from using third-party AI agents to book flights or access MileagePlus accounts, citing privacy risks and possible conflicts with FAA and DOT rules. The decision, reported by the New York Times and Retail Customer Experience, highlights where AI agents should be permitted to act for customers.
What happened
United Airlines has barred customers from using third-party AI agents to book flights or access their MileagePlus accounts, according to reporting by the New York Times and Retail Customer Experience. The airline cites privacy risks and the possibility that automated actions could break federal aviation rules.
United's concern centres on credential sharing. Handing login details to a personal or third-party AI assistant could expose sensitive data, and an agent acting without a human involved might make unintended purchases, change account settings or spend miles. The airline also argues that automated bookings could conflict with FAA and DOT requirements, which in some cases, such as the check-in declaration that bags contain no banned items, expect a person to give the confirmation.
The decision comes as consumers rely more on AI tools for shopping, travel planning and advice. Retail Customer Experience reports that 70% of consumers now use AI in their personal lives, which makes the airline's position a visible test of how far AI agents should reach into transactional services.
Why it matters
This is less a story about one airline than about where AI agents can safely act on a customer's behalf. Booking, rebooking and account management are high-stakes, identity-linked transactions, and United's ban signals that operators may treat delegated AI access as a security and compliance question first and a convenience feature second. For leaders in digital transformation, the practical implication is that agentic capability needs clear permission boundaries, audit trails and a defined point where a human confirms the action.
For the customer, the experience gap is one of trust and control. A traveller who wants an AI assistant to handle a disruption at 11pm needs to know exactly what it can change, what it cannot, and how to undo a mistake. Organisations that deploy AI in service channels will need to design those guardrails explicitly, rather than assuming customers will accept whatever an agent does with their credentials.
The Renascence take
Most coverage will frame this as a pushback against AI. The more useful reading is that United has drawn a line between assistance and delegation, and customers are unlikely to see that distinction clearly unless operators make it visible.
The real lesson is that an AI agent is only as trustworthy as the permissions it has been given and the moments it must stop and ask. Restricting access is a blunt instrument; the better design principle is to let AI prepare, recommend and draft while a person confirms anything irreversible. Customer-obsessed operators should map every AI-accessible action by its reversibility and financial consequence, then design explicit consent and undo steps for each. If you cannot explain to a customer exactly what your agent can do with their account, it is not ready to do it.
Sources
This briefing was written by our Newsdesk, synthesising reporting from the outlets below. Follow the links for the original coverage.
FAQ
Questions we get on this topic
More in Hospitality
Stay ahead of CX
Get the signal, not the noise.
The stories shaping customer experience — plus the Journal and Experience Loom — in your inbox.


