A post meant to showcase an AI shopping assistant also showed how much purchase information can be exposed in a screenshot. Elon Musk shared details of an order for Lego Star Wars kits on X, and other users soon posted claims that they had used the information to interact with Lego’s customer service chatbot.
The screenshot contained more than a shopping demonstration
Musk had been describing how useful his Grok Bot AI agent was for placing an order. The screenshot he shared with his more than 240 million followers included a 10-digit Lego order number, the amount paid for the sets and delivery, billing and delivery addresses, the last four digits of a credit card, and the email address the bot used to buy the bricks.
Those details gave the post a consequence beyond showing that an agent could help with shopping. The order number and email address were among the information another person could use when contacting Lego’s online assistant about the purchase.
X users posted claims about changing the order
An X account holder named Enderman shared screenshots of interactions with Lego’s customer service chatbot. The user appeared to use Musk’s email address and order number to ask the system to cancel the order.
Musk replied within a couple of hours that the order had not been canceled. That response left open whether the chatbot interaction had worked, but did not end the attempts to use the publicly posted information.
Another X user later shared screenshots suggesting they had used the same details with Lego’s AI assistant, called “Sophia.” The screenshots appeared to show an attempt to reopen the delivery, increase the order count to 400, and change the delivery address while keeping Musk’s original payment method. Sophia reportedly replied, “Perfect! I’ll get your Lego adventure back on track.”
The screenshots describe what users said they were able to do. Lego declined to comment when WIRED contacted the company to confirm whether this activity took place on Musk’s personal Star Wars order.
What the episode shows about AI shopping
Musk had suggested on Saturday that Grok could make purchases online from a simple prompt. In a post, he said users could upload a picture of their credit card so the bot could search for deals and complete a purchase. The Lego order became a public example of the risks that can accompany this kind of convenience: a shopping assistant may act on a customer’s behalf, while information about the resulting order can still be exposed in a post.
Other AI companies, including Meta and OpenAI, have also highlighted shopping tasks as a use for their agents. The source describes Visa’s separate warning about fraud in agentic commerce. Visa said bad actors were building legitimate-looking sites that could mislead AI agents into buying fake items by offering goods at very low prices.
Visa’s blog warned, “Once the AI agent completes the purchase … the malicious merchant can harvest payment data and instantly use it for unauthorized transactions.” It also said it observed a 25 percent increase “in malicious bot-initiated transactions” in the latter half of 2025, with the figure reaching 40 percent in the US.
Order safeguards still matter
The Lego episode raises a practical question about how a chatbot handles a request to alter an order when someone has access to its identifying details. The source notes that Lego’s AI assistant could have internal safeguards capable of flagging unusual activity, and says human staff may have been tasked with checking changes. It does not establish whether those safeguards were triggered or whether any attempted changes took effect.
For customers using AI to shop, the incident underscores the value of treating order screenshots as sensitive. A receipt or confirmation screen can contain details that make it easier for someone else to contact a retailer about a purchase. And when an AI assistant can place or manage orders, the retailer’s process for checking who is making a change remains part of the security picture.
The reported attempt to turn one order into 400 sets also illustrates how a seemingly routine customer service interaction could have significant consequences if a request were accepted without adequate checks. In this case, the available account consists of screenshots posted on X and Musk’s reply; Lego did not confirm the activity to WIRED.