Letting an AI agent spend your money is a very different kind of trust from asking it a question, and that difference was on full display on X on Thursday, October 15.
Elon Musk replied to a user who said Grok Bot had booked a flight for him after he shared a photo of his payment card, calling the process far simpler than working through an airline website. In his reply, Musk said SpaceX would make a user whole if the bot made a mistake with a purchase or leaked card details, adding that a leak has never happened yet. It was a public assurance made in a post on X, not a detailed new policy document.
Hours later, the limits of agent shopping became just as clear. Another user shared an Amazon error saying continued access by an unauthorized AI agent violated the site's Conditions of Use, and said his bots could no longer place orders there. Musk's advice was to ask Grok Bot to shop somewhere else instead, arguing there are many purchasing sites beyond Amazon. After a user suggested an X Shop, he floated the idea of starting a competitor to Amazon.
That exchange should be read as a tease, not a launch. Reporting on the posts noted no confirmation that a rival store is being built, and Amazon had not commented publicly in the coverage reviewed. The block itself was not entirely new either: users had raised Amazon blocking the bot as early as October 10, when Musk responded that Grok would find another way and pointed to an order placed through Lego's website.
For X users, some background helps explain why this matters now. Grok Bot is described in the reporting as a set of autonomous agents that can handle multi-step tasks such as shopping, booking and administrative work. It arrived in early beta in August and is available through certain Grok subscription plans, including SuperGrok Heavy. In other words, it is an early product being tested in the open, on real transactions, by real users posting their results.
There is also an important gap between a promise in a post and the written terms that govern an account. Teslarati noted that a similar make-you-whole assurance was offered in late August in connection with a finance integration, and that the consumer terms it reviewed still place responsibility for an agent's actions on the user, with liability generally capped. Neither of Thursday's posts said those terms had changed. Until a formal policy with eligibility and claim steps is published, readers should treat the assurance for what it is: a notable statement from Musk, with its scope still unconfirmed.
The practical takeaway is caution paired with curiosity. If you try an agent to book or buy, start small, review every confirmation yourself, and avoid handing over payment details you are not comfortable sharing. Keep screenshots and receipts, and check the retailer's own rules, because a site can decline agent access under its terms, as Amazon has done here. The idea of shopping by conversation is clearly moving forward on X. Thursday showed both its appeal and the questions that still need written answers.