Elon Musk Exposes Purchase Credentials in Grok E-Commerce Demo, Raising Agentic AI Security Concerns

xAI founder Elon Musk inadvertently exposed sensitive transaction details on social media platform X after sharing a screenshot demonstrating the automated purchasing capabilities of his company's Grok AI bot. Musk, who has over 240 million followers on the platform, posted the image on Sunday to showcase how the AI agent ordered Lego Star Wars sets on his behalf.

The published screenshot revealed critical purchasing details, including a 10-digit Lego order number, the total purchase price, billing and delivery addresses, the last four digits of Musk's credit card, and the email address used by Grok to complete the purchase from Lego.

Following the post, public users on X attempted to manipulate the transaction through automated customer service tools, as first reported by Wired (https://www.wired.com/story/elon-musk-bragged-about-his-grok-lego-star-wars-haul-then-x-users-punked-him/). An X user under the handle Enderman posted screenshots showing an interaction with Lego's website chatbot, using the exposed email and order number in an attempt to cancel Musk's order. Musk replied within two hours, stating that his order had not been canceled.

Subsequent attempts were made by another user using Lego's customer service AI assistant, named “Sophia.” According to screenshots shared on X, the user submitted the exposed details to reopen the delivery, expand the quantity to 400 sets, and alter the delivery address while keeping Musk's original payment method. The chatbot reportedly confirmed the modification during the session.

Lego declined to comment when contacted by Wired to confirm whether the reported activity occurred on Musk's order.

The incident followed remarks Musk made on Saturday suggesting that Grok could execute e-commerce purchases from a single prompt, proposing that users upload a photo of their credit card so the model can search for deals and complete transactions. Other artificial intelligence developers, including OpenAI and Meta, have also promoted autonomous shopping agents.

The disclosure highlights emerging operational and security risks in agentic commerce. In an advisory published late last year, Visa warned of fraud in AI-driven commerce, noting that malicious actors create fake storefronts with artificially low prices to trick shopping bots into harvesting payment information for unauthorized transactions. Visa reported a 25 percent increase in malicious bot-initiated transactions globally in the second half of 2025, with that figure rising to 40 percent in the United States.