Elon Musk Showed People How To Shop With Grok Using Credit Cards: Then He Exposed His Own Order Details

Elon Musk has been promoting Grok Bot as an artificial intelligence agent that can handle online shopping, but his attempt to demonstrate the technology has drawn attention to a basic problem with automated purchases: how much personal information should users be willing to expose?

The billionaire posted a screenshot of an order for LEGO Star Wars sets after showing followers how Grok could find products and buy them.

The image revealed details that could be used to identify and access the purchase, prompting other users to test what they could do with the information. One claimed to have submitted a cancellation request through LEGO's customer service chatbot, while another posted screenshots purporting to show changes to the order.

Neither account established that the changes were completed. The episode offers a timely warning as AI companies develop tools that can act on users' behalf, rather than simply answer questions. Giving an agent purchasing power also raises questions about privacy, authorisation and who remains responsible when something goes wrong.

Elon Musk's Grok Bot Shopping Demo Exposes Personal Details

On 10 October, Musk posted on X that users could photograph their credit cards and upload the images to Grok Bot, which would search online for the best price and place an order.

'To buy stuff, you can just take a picture of your credit card, drop it in the chat and Grok @Bot will scour the Internet for the best deal on the item you want and order it,' he wrote.

The following day, Musk shared a screenshot of Grok Bot helping him select LEGO Star Wars kits. The displayed order confirmation showed five sets costing $241.34 (approximately £180), with free delivery, according to the screenshot reported by WIRED.

An X account named Enderman subsequently posted screenshots that appeared to show LEGO's customer service chatbot, Sophia, accepting the order number and submitting a cancellation request. The exchange did not establish that the order had actually been cancelled.

Musk later responded on X, insisting that the order remained intact. LEGO declined to confirm whether any cancellation or other changes had taken place per reports

Another user shared screenshots allegedly showing the chatbot increasing the order to 400 sets and changing its destination to Little St James, the island associated with convicted sex offender Jeffrey Epstein. Those claims also remain unverified, and the posts alone do not prove that the additional purchases or address change went through.

AI Shopping Agents Raise Credit Card Security and Privacy Risks

AI agents can browse websites, compare products and take actions on a user's behalf, but those capabilities also create new ways for misleading information to influence a transaction.

In a November 2025 security analysis, Visa warned that fraudulent retailers could create convincing websites designed to mislead AI shopping agents. Such sites may offer unusually low prices to encourage purchases, potentially exposing payment information if a transaction succeeds.

This is a separate risk from the LEGO incident, in which the reported concern centred on order details displayed in a public screenshot.

OpenAI has also outlined the risks of prompt injection, in which malicious instructions hidden in web pages, emails or other content attempt to manipulate an AI system into doing something the user never requested.

The company advises users to give agents specific instructions, monitor their activity on sensitive websites and check consequential actions before approving them.

For shoppers, safeguards around an AI purchase matter just as much as convenience. A system should make clear what information it needs, which transactions it can authorise and when a person must approve a payment or change to an existing order.

Users can reduce their exposure by avoiding public screenshots of receipts and order confirmations, limiting the financial permissions granted to automated tools, and requiring approval before purchases are completed.

They should also review the information an agent shares with third-party websites and avoid uploading card images unless the service's handling of that data is clear.

Musk's demonstration showed how quickly an online shopping example can become a privacy concern. The unverified LEGO claims do not prove that Grok itself was compromised, but the exposed order details illustrate why AI shopping agents need clear limits on access to financial information and safeguards around changes to purchases.