X Users Exploit Elon Musk Grok Lego Order Data
X users exploited personal details in a screenshot shared by Elon Musk showing his Grok AI buying Lego sets, highlighting critical security gaps in autonomous e-commerce agents.
Elon Musk recently demonstrated how his Grok AI agent could make online purchases, but publishing the transaction proof triggered an immediate security lesson for automated shopping systems. After using Grok to order Lego Star Wars products, Musk shared a screenshot on X with his more than 240 million followers. The image inadvertently contained a 10-digit Lego order reference number, full delivery and billing addresses, his bot's email address, and the final four digits of his payment card.
Social media users quickly tested the vulnerabilities of traditional customer support channels using the exposed details. An X account named Enderman submitted the information to Lego's automated web chat assistant, named Sophia, in an attempt to cancel the purchase. Though Musk claimed the order survived, another user subsequently leveraged Sophia to alter the order destination, increase the item count to 400 sets, and bill Musk's existing payment method.
The incident underscores growing risks in agentic commerce as tech giants promote AI shoppers that handle financial credentials directly. Musk previously suggested that users upload credit card images to Grok so the assistant could scan for deals and complete checkout processes automatically. However, financial institutions have raised alarms regarding automated fraud. Visa reported a 25 percent increase in malicious, bot-initiated transactions globally during the second half of 2025, with US attacks spiking by 40 percent as bad actors create fraudulent storefronts to deceive shopping bots into leaking payment data.
For developers and commerce platforms, the exposure illustrates that autonomous agent workflows require stringent identity verification beyond basic order numbers. When automated agents like Sophia process high-value adjustments or order modifications without secondary authentication, bad actors can exploit valid AI credentials to alter fulfillment pipelines, risking scenarios far more severe than unauthorized toy orders.
This is our own summary of reporting by WIRED AI
