United Airlines blocks customer AI agents, citing FAA and data risks

United Airlines, the world's second-largest carrier at $59 billion, on Wednesday became the latest enterprise to block customer-facing AI agents — and it did so in writing, on the record, with an appeal for others to follow.

The airline said the move responds to fears of data leakage, unintended purchases and violations of federal airline rules. It follows similar restrictions from Walmart, Target and Amazon. For CIOs watching the dominoes fall, the question isn't really whether more bans are coming. It's what, if anything, would make these companies open the door again.

What exactly did United decide?

United said it is restricting access "directly within our digital platforms and indirectly via travel agencies and aggregators" to prohibit unauthorized third-party personal AI agents or assistants, including updates to its booking, ticketing and distribution policies.

The airline's concerns are concrete:

• Customers sharing account credentials with AI assistants "can expose sensitive information, and actions taken by these tools may result in unintended purchases, account changes, or use of MileagePlus miles."
• Automated AI actions "have the potential to violate FAA and DOT regulations" — such as rules requiring a human to acknowledge at check-in that bags contain no banned substances like explosives.

United also framed the ban as leadership, noting that banking and retail executives have taken similar actions, and said it "encourage[s] others to consider measures that preserve the data and security of their customers and strengthen safety." Crucially, it left the door cracked open: "As personal AI assistants evolve, United will continue to evaluate how these tools interact with our services, with a focus on protecting our customers."

How are Walmart and Target handling it?

The retailers diverged sharply. Walmart, in a program scheduled to start this month, is trying to separate bot activity from human activity. "Today, when a customer uses an AI agent, it signs in using a customer's account. This means that all purchases — whether made by the customer or the agent — appear together, without any differentiation," Walmart's announcement said. Its new Agent Profiles will let customers authorize each AI agent with a dedicated profile, so agent purchases appear separately.

Target went the other way. Its updated terms and conditions block all but a handful of approved agents — its own AI Shopping Assistant plus OpenAI, Google and Microsoft — and it "may add, remove, or otherwise modify the list of Approved Agents at any time and in its sole discretion." Every other agentic commerce tool is "unauthorized, expressly prohibited" from accessing, logging into, scraping or querying the site "under any circumstances," whether directly or through API calls. Approved agents can't even be combined with unapproved ones without Target's written consent.

Why are enterprises saying no?

Industry observers point to a deeper problem than any single security risk: there's no standardization around AI agents, and the number of bots from major AI vendors and startups keeps growing. Supporting every bot customers use could become an almost infinite drain on IT resources — though strictly enforced industry standards, they say, might persuade some enterprises to reconsider.

Frank Dickson, principal analyst at Dickson Research, put it bluntly: "Enterprises do not fear bots. They fear liability that nobody has assigned. Once a rule says which steps a human must perform, which steps an agent may perform, and how each is recorded, a general counsel can sign off on the deployment. Ambiguity is what keeps bots stuck in pilot."

He argued the bans relocate value rather than destroy it. "Blocking does not destroy the value of agents. It moves it. Every wall eventually gets a gate. The gate charges admission." United's own wording — prohibiting only "unauthorized" agents — leaves room for sanctioned ones. "Whoever owns the agent owns the customer relationship," Dickson said. "Airlines have spent years pulling customers back from third-party travel sites. They are not going to hand that relationship to someone else's chatbot."

The operational math is brutal, too. "A sea of agents is a sea of requests. Somebody pays for every one," Dickson said, noting each agent arrives with its own login method, its own idea of customer consent and its own activity record — if it keeps one at all. "Bots that check prices over and over do not buy more seats."

What does this change for buyers?

Noah Kenney, principal consultant at Digital 520, sees no downside for United: "substantial risk, an uncertain commercial upside, and almost no competitive pressure to accept either" lets the airline simply wait until the legal consequences of bot behavior become predictable.

His broader worry is scope. "One of the biggest mistakes I see in enterprise AI deployments is giving agents access to far more information and systems than their responsibilities actually require," he said. "There is a fundamental difference between an agent that is instructed to access only one system and an agent that is technically incapable of accessing anything else."

Justin Greis, CEO of consulting firm Acceligence, expects more restrictions while companies build controls — while cautioning that "widespread blocking would reduce agents' value by preventing them from completing tasks across services." He also sees a training gap: "actions they authorize through an agent can still create obligations for them, even when they never personally click 'buy,' 'send' or 'confirm.'"

For now, the wall is up. The gate, everyone agrees, is still being designed.