Musk's own screenshot of a Grok-placed Lego order showed addresses, card digits and an order number, and X users then used Lego's chatbot to meddle with it.
Elon Musk showed off his Grok AI agent by posting a screenshot of a $241.34 Lego Star Wars order it placed for him. The image, seen by his 240 million-plus X followers, included the order number, a delivery and a billing address in Texas, an email and the last four digits of his card.
X users then took those details to Lego's support chatbot. Screenshots appear to show it agreeing to cancel the order, and later to raise it to 400 sets with a new address. Musk said the order went through, and Lego declined to comment.
The lesson for everyone else is simple. Musk had also suggested uploading a photo of your credit card so Grok can shop for you, even though xAI warns against sharing personal data in prompts. Crop order details out of screenshots, and keep card photos out of chatbots.
Elon Musk wanted to show how easy it is to let an AI agent go shopping. Instead, the screenshot he posted to more than 240 million followers on X showed his Lego order number, the price, two addresses, an email and the last four digits of a payment card, and strangers quickly used those details to tamper with the order.
The episode is funny on the surface, since the order was five Lego Star Wars sets. But it lands right as tech companies push AI agents that can spend your money. If the owner of the AI company can leak his own order details in one post, it’s worth asking what ordinary users might expose, and how easily a store’s own chatbot can be talked into changing an order.
The Grok Lego order leak started with a boast. Over a weekend in mid-October, Musk wrote that he had asked Grok to find the Lego Star Wars sets that small children would enjoy most and buy them, adding that it was that simple. He attached a screenshot of the finished purchase on Lego’s website.
That image showed five items totaling $241.34, a 10-digit order number, a delivery address in Austin, Texas, a separate billing address, the email address the bot used to buy the sets, and the last four digits of an American Express card. A cybersecurity account on X reposted it with a sarcastic note about broadcasting a home address to millions of people.
Musk replied that the Austin address was where he gets his mail, not where he lives. A few days later he said the sets had arrived, though he shared screenshots of his Grok chat rather than photos of the boxes.
This is where it turned from embarrassing to instructive. An X user called Enderman posted screenshots of a chat with Lego’s customer service bot on its website. Using only the email and order number from Musk’s post, the bot appeared to agree to cancel the order.
Within a couple of hours Musk said the order had not actually been cancelled. Another user then went back to Lego’s AI assistant, named Sophia, and appeared to reopen the delivery, raise the quantity to 400 sets and switch the shipping address, all while keeping Musk’s original card on file. The screenshots show the bot cheerfully promising to get the Lego order back on track.
None of this has been confirmed by Lego, which declined to comment. It’s possible that internal checks flagged the odd changes and that a human reviewed them before anything shipped. But the screenshots suggest an order number plus an email was enough for the chatbot to accept changes.
That is the bigger lesson for stores. Many retailers now put AI assistants in charge of order changes, and those assistants are only as careful as the checks behind them. If knowing two pieces of information printed on a receipt is enough to redirect a parcel, anyone who sees that receipt can try it.
I just asked Grok @Bot to research the most fun Lego Star Wars kits for young kids and order them. As easy as that. https://t.co/ZWqEIwBNFB pic.twitter.com/5IKVoqvuhb
— Elon Musk (@elonmusk) October 11, 2026
The Lego post came a day after Musk suggested the simplest way to shop with Grok: take a photo of your credit card, drop it into the chat, and Grok will hunt down the best price and place the order.
That advice clashes with xAI’s own guidance, which tells users not to include personal information in prompts. Under xAI’s privacy policy, the company can keep content people upload, including images, and use it to improve its services and train its models. Deleted chats are generally meant to be removed within 30 days, with some exceptions.
It’s also unclear how independent the purchase really was. Musk’s screenshot doesn’t show whether Grok Bot finished the checkout by itself or whether he approved the payment.
Grok is not alone. Meta, OpenAI, Google and Anthropic have all promoted AI agents that can browse sites and complete tasks for you, shopping included. That convenience comes with two kinds of risk.
The first is what you share. A card photo in a chat window is a copy of your card that now lives on someone else’s servers. The second is what agents can be tricked into doing. Late last year Visa warned that scammers were building convincing fake shops with very low prices to lure AI shoppers, then collecting the card details and using them for fraud. Visa said malicious bot-driven transactions rose 25 percent in the second half of 2025, and 40 percent in the US.
A few habits help:
Musk said it was not, and he later posted that the sets had been delivered. Lego declined to comment on any changes made through its chatbot.
An order number, the total paid, a delivery address and a separate billing address in Texas, the email used for the purchase and the last four digits of a card.
xAI itself tells users not to put personal information in prompts, and its policy lets it keep uploaded images, so it’s best avoided.
You're reading the quick version.
Read next
Paying $99.99 to vanish from people-finder sites? Malwarebytes, rated
It finds you on people-finder sites and asks them to delete you. But the price, broker count and refund rules depend on where you look.
About 11,000 W5 PRO chargers sold on Amazon are affected. Here's how to tell in ten seconds, how the refund works and how to get rid of it safely.
The crypto-tracing firm says an attacker used AI agents to reach its internal tools. Here's what was taken, who's exposed and what users should do now.
A swapped page-speed script drained 726 wallets in under an hour. Here's what happened, who is being paid back, and how to protect your account.
The Daily Brief
Every morning, the news that matters from AI, phones, apps and the people shaping tech. Explained in plain English. Free.
One email a day. No spam, unsubscribe anytime. Privacy policy
