Newsletter

United restricts unauthorized AI assistants: according to United, a human must confirm the baggage declaration

United restricts unauthorized AI assistants. One colleague tries AI, another waits. And Adobe Express from inside your AI chat.

By Piet Baudoin · 12 November 2026

United Airlines is restricting access for third-party AI assistants it has not authorized, and gives as an example that under the rules a human has to confirm at check-in what is in the bags. Also: the colleague who experiments with AI next to the colleague who waits, and designing from your AI chat.

Can an AI assistant book my flight and check me in?

On 11 November, United announced that it is restricting access for “unauthorized, third-party personal AI agents or assistants”, including through travel agencies and aggregators. First reason: if you share your login details with such an assistant, you can expose sensitive information, and the assistant may make unintended purchases.

Second reason: such actions can violate US aviation rules. The example United gives is the most interesting sentence in the release: “flight check-in requires human acknowledgment that bags do not contain banned substances like explosives.” An AI assistant can arrange almost an entire trip; United says a human must confirm that statement about the suitcase. According to United, corporate leaders in banking and retail have taken similar actions. The release gives no start date.

So if you let an AI assistant arrange travel, the airline has a say in which assistant is allowed. And you don’t simply hand an assistant your password.

Source: United, 11 November 2026

Why does one colleague try AI while another waits?

Derwish Rosalia, a Dutch chartered accountant and trainer, describes in Accountancy Vanmorgen, a Dutch trade magazine for accountants, a difference between colleagues: “One colleague experiments with AI extensively, while the other waits” (translated from Dutch). Don’t label that difference as resistance straight away, he says: first ask what someone needs to try one well-defined task responsibly. Rosalia also teaches the AI courses the magazine offers.

Doubts about quality are sometimes simply justified. The model that searched the news for this newsletter this morning first picked a supplier’s customer story as a news item; the magazine itself labels it as Branded Content. It was dropped on review.

A day later, the same magazine cited the employer branding research Randstad published in May, with 13,430 respondents in the Netherlands: “31% experience no say at all in the introduction of AI in the workplace” (translated from Dutch).

Who is still waiting at your company, and has anyone asked why?

Source: Accountancy Vanmorgen, Derwish Rosalia, 10 November 2026 · Accountancy Vanmorgen on Randstad, 11 November 2026 · Randstad, 28 May 2026

Can I create a design from my AI chat and keep working on it myself?

On 11 November, Adobe described how Adobe Express, its app for posts and flyers, works inside ChatGPT, Claude, Microsoft Copilot and Gemini. Part of this already existed; this post puts all four side by side. You ask for a design in the chat and can continue in Express yourself at any moment: “your changes sync both ways”. What is available varies by chat app. Adobe sells the software itself.

Source: Adobe Community, 11 November 2026

Where are we heading?

United names two objections: an assistant with your login details poses as you, and some steps, such as the statement about the bags at check-in, require confirmation by a human, according to United. A shared password does not by itself prove who confirms that statement. I expect that within five years passwords will no longer be needed: you prove who you are by being recognized, for example by your voice or your face. Being recognized is not yet a yes. So you still confirm the statement about your suitcase yourself, with one glance at your phone at the moment it matters. For that, companies need to be able to link the permission for that one action to the traveler. Your agent then finds and books the flight, and companies like United build their own entrance for their customers’ agents, where that link is visible.

What does this mean for your business?

At United it is about the bags; at the office it is about email: which messages may an AI agent prepare, and who gives permission to send them? That is also the subject of our piece on whether an AI assistant should send email on its own.

With Bombos, an AI agent drafts the replies in the shared inbox; a staff member reviews and sends them. That one task is the start; after that, your team teaches Bombos the next one. The free scan shows which task you can start with.

Sources

Every source was opened on 12 November 2026; every quote appears in it word for word, quotes from Dutch sources in our translation.

United Restricts Access for Third-Party AI Agents (11 November 2026) · Blog | AI-tip: Maak van terughoudendheid een concrete leervraag (10 November 2026) · Bedrijven tasten flink in de buidel voor AI-talent (11 November 2026) · Randstad: Gebrek aan invloed op AI baart werknemers meer zorgen dan baanverlies (28 May 2026) · Design in ChatGPT, Claude, Copilot, and Gemini — Adobe Express Is Now in Your AI Chat (11 November 2026)

The first half hour · free

What does your AI forget?

  1. 01You tell us which AI you use and where it keeps starting over.
  2. 02We show what a shared memory changes there, on your own examples.
  3. 03You keep the plan, with or without Bombos.
Piet Baudoin
Piet BaudoinYou talk to Piet, from the Bombos teamDutch team from Utrecht and Nijmegen · data in Europe
We reply the same working day.