Musk's own X screenshot leaked order details that let strangers hijack his Lego order via chatbot

Elon Musk recently posted a screenshot on X of an AI-assisted purchase, and the image carried more than product details: it showed a 10-digit order number, a full delivery address and partial payment information. According to a report on dev.to, originally published by Hacks.gr, that was enough for other users to act on the order through the vendor's automated customer service system — a stray screenshot turned into a working account takeover.

What the screenshot exposed

The purchase was made through an AI assistant from Lego, and the shared frame included the order number, the full address and partial payment data. The report's central observation is that the interface displayed state identifiers — values that effectively act as proof of ownership — inside a UI element that people routinely share in public posts. Once posted, those identifiers stopped being private.

How strangers moved the order

An X user took the exposed email address and order number into Lego's automated customer service system and, according to the report, was able to cancel the order. A few hours later, Musk tweeted that the order had not been cancelled after all. Another user then allegedly used the same details in the same system to reorder the delivery, raise the quantity to 400 sets and send it to a different address. When reporters contacted Lego to confirm what had happened, the company declined to comment.

Warnings that came first

The episode landed shortly after Musk suggested that X users let digital assistants handle online purchases by uploading a photograph of their credit card. Visa, according to the report, had been warning since late last year about fraud risks in AI-driven automated shopping and about a rise in transactions initiated by automated programs with malicious intent.

Why it matters

Nothing in this incident required a sophisticated attack. The only credentials involved were an order number and an email address visible in a public post, yet the chatbot accepted them as sufficient to cancel and then rewrite an order. That is the systemic risk the report points to: when AI agents and automated support are grafted onto legacy customer service systems, impersonation gets cheaper, because authentication rests on knowledge of a transferable identifier rather than on verifying the actual account owner.

It also matters because agentic shopping is the direction of travel. As assistants place orders and manage them on a user's behalf, identifiers like order numbers behave like bearer tokens, and a screenshot becomes the equivalent of a leaked password. Vendors rolling out chatbot-based support need step-up verification before allowing sensitive changes such as new delivery addresses, larger quantities or payment edits — and anyone sharing a purchase screenshot should assume every number in the frame is a secret.