In a recent display of overconfidence regarding the purchasing capabilities of his Grok AI, Elon Musk shared a screenshot of his Lego Star Wars order with his 240 million followers on X, inadvertently exposing sensitive personal and financial information. The image revealed his 10-digit order number, the cost of the sets, billing and delivery addresses, the last four digits of his credit card, and the email address used for the transaction.
Within hours, an X user identified as Enderman leveraged these details to interact with Lego’s customer service chatbot, successfully canceling Musk’s order. Although Musk later claimed the purchase remained active, the incident highlighted significant security vulnerabilities in using AI agents for e-commerce.
Another user subsequently exploited the same information through Lego’s AI assistant, Sophia, to reopen the delivery and increase the order quantity to 400 sets while retaining Musk’s original payment method. According to posted screenshots, the AI confirmed the changes with the message, “Perfect! I’ll get your Lego adventure back on track.”
This event follows Musk’s earlier promotion of Grok’s ability to autonomously shop online by uploading credit card images to search for deals and finalize purchases. Other tech giants, including Meta and OpenAI, have similarly emphasized the shopping utility of their respective AI agents.
However, Visa issued a warning late last year about fraud risks in agentic commerce, noting that malicious actors were creating fake websites to deceive AI into purchasing counterfeit goods. The company reported a 25 percent rise in malicious bot-initiated transactions globally in the second half of 2025, with the figure reaching 40 percent in the United States.
Lego declined to comment on whether internal safeguards flagged the unusual activity surrounding Musk’s order, but it is likely that human oversight intervened to review the alterations given the potential for AI error.
Visa warned about this exact thing last year. It’s alarming how few companies actually have safeguards against AI agents being manipulated by bad actors.
Wait, they increased the order to 400 sets? I hope Elon actually bought them all for charity. That would be the ultimate flex.
Is it really Grok’s fault though, or just Musk’s poor operational security? Sharing a full receipt with everyone seems like a rookie mistake.
This is a hilarious but terrifying lesson. You literally cannot share order details publicly if you want your purchase to remain intact.