United Airlines Restricts Unauthorized AI Agents From Booking and Changing Trips
United is the first U.S. airline to say explicitly that it will block personal AI assistants from completing bookings, citing account security, unintended purchases and passenger acknowledgment rules. Agents can still search fares.
By The AI Enterprise Magazine Team3 min readNews · Weekly briefing

Key takeaways
- Transactions blocked. Unauthorized agents may not book, change or service tickets or buy products, on United channels or via agencies and aggregators.
- Search still allowed. Assistants can surface fares but cannot complete the purchase.
- Stated reasons. Credential sharing risk, unintended purchases and FAA and DOT acknowledgment rules.
- Two models emerge. Delta reportedly allows authenticated assistants, showing an alternative to outright blocking.
United Airlines said on November 11 that it will restrict unauthorized third-party AI agents from booking and managing travel on customers' behalf, becoming the first U.S. carrier to state such a policy explicitly, according to Skift. The restriction applies on United's own digital platforms and indirectly through travel agencies and aggregators. It is an early, concrete example of a large company deciding that outside agents acting with customer credentials are a risk it will not accept by default, and it gives other enterprises a template for thinking about agent access to their own systems.
What exactly is United restricting?
United is updating its booking, ticketing and distribution policies to bar unauthorized personal AI agents and assistants from completing transactions, Digital Trends reported. InsideFlyer reported that the covered actions include making, modifying or servicing a ticket and buying any service or product, and that the scope extends to autonomous and semi-autonomous assistants, including supervised actions. United added the ban to its list of prohibited ticketing practices.
The policy does not stop AI tools from finding flights. Digital Trends, citing The Wall Street Journal, reported that assistants can still surface United's flight options and fares but will not be allowed to complete a booking. InsideFlyer noted that fare alerts remain possible, but an assistant cannot act on a lower price it finds. None of the reports described a technical enforcement method or a specific effective date, and Skift said United would evaluate the policy as personal AI assistants evolve.
Why did United say it is doing this?
United's stated reasons fall into three groups. The first is security and privacy. United said that third-party AI access to customer accounts introduces significant security risks, and Skift reported that sharing account credentials with such tools can expose sensitive information. The second is unintended transactions, including purchases, booking changes and MileagePlus redemptions that a customer did not mean to make.
The third is regulatory. Some steps in air travel require the passenger's own acknowledgment, such as confirming at check-in that bags contain no prohibited or hazardous items. United argued that an agent completing those steps automatically could conflict with FAA and Department of Transportation requirements, according to Skift and Digital Trends.
There is also a commercial dimension. Skift and Digital Trends both reported, citing The Wall Street Journal, that United CEO Scott Kirby said he did not want to create another middleman between the airline and its customers.
How are other companies handling agents?
United's approach is not the only one. Digital Trends reported that Delta reportedly allows properly authenticated AI assistants to make reservations with extra safeguards, and that Southwest has a ChatGPT integration for flight search. The same report noted that Amazon and several major banks have restricted unauthorized AI agents, while Google and OpenAI continue to build agents that handle travel tasks.
The split matters because it shows two models emerging. One treats outside agents as a threat to be blocked unless explicitly authorized. The other builds an authenticated path for agents so that the business keeps control of identity, consent and the transaction record.
What does this mean for enterprise leaders?
For companies that run customer-facing digital channels, United's move raises a question many have not answered yet: what is the policy for AI agents that log in as customers? Agents that use a customer's username and password look like that customer to most fraud and authentication systems. If an agent makes an unwanted purchase, the dispute, the refund and the reputational cost land on the business, not on the agent's developer.
Three practical steps follow. First, inventory where agent traffic already reaches customer accounts, since many organizations will find it in logs before they have a policy. Second, decide which actions require a human acknowledgment, as United did for regulated check-in steps, and design those steps so an agent cannot complete them silently. Third, consider whether an authenticated agent channel, with scoped permissions and clear consent records, would serve customers better than an outright block.
For enterprise travel and procurement teams, the policy has a more immediate effect. Corporate deployments of AI travel assistants that book on United for employees may need to route through authorized channels or stop at fare search. Travel managers should check how their booking tools and agencies are interpreting the new rules, since United said the restriction applies through agencies and aggregators as well as its own platforms.
Frequently asked questions
Can AI agents book United Airlines flights?
Not unauthorized third-party agents. Under a policy announced November 11, 2026, they cannot complete bookings or service tickets, though they can still surface fares.
Why is United blocking AI agents?
United cites security risks from shared account credentials, unintended purchases and MileagePlus redemptions, and rules requiring passengers to acknowledge items such as hazardous materials themselves.
What should businesses learn from United’s AI agent policy?
Decide which customer actions require a human acknowledgment, monitor agent traffic on customer accounts, and consider an authenticated agent channel instead of credential sharing.
Sources
Filed under: Agentic AI, Security & Fraud, AI Governance & Trust
Become a Sponsor







