Elon Musk's AI agent demo accidentally leaked his own personal data.
Musk wanted to show off Grok's new agentic abilities — the AI doing a multi-step task instead of just chatting. He had it place an order for Lego, then posted a screenshot to X to prove the purchase went through. The image showed more than a checkout confirmation. It also included personal information that had no business being public, posted straight from his own account.
This isn't really a Lego story. Screenshots capture everything on screen, not just the part you meant to show, and that gets riskier as "agent" tools start completing real transactions instead of just answering questions. Every demo of an agent clicking through a live checkout risks exposing addresses, saved cards, or account details meant for one pair of eyes only.
The irony: the person running the company that built the agent is the one who got caught by its own screenshot.