xAI Template HTML Injection Phishing Scam

A sophisticated phishing campaign recently targeted high-profile individuals by exploiting a vulnerability in the xAI platform's email invitation system. The attack involved naming an xAI team with raw HTML code, which was subsequently rendered by the platform's invite template without proper escaping. This allowed a malicious 'Accept Invitation' button to appear within a legitimate email sent from the 'noreply@x.ai' domain.

Technically, the attack succeeded because the phishing emails passed all standard authentication checks, including SPF, DKIM, and DMARC, since they originated from xAI's authorized infrastructure. While the Grok Bot AI agent was successfully used to identify the hidden link, security researchers warn that using AI agents for such tasks carries significant risk. Because these agents often run on shared cloud resources within a single account, a malicious prompt in one bot could theoretically influence other bots on the same network, potentially exposing financial data or credentials.

This incident highlights a growing trend where attackers exploit the legitimate workflows of AI service providers to bypass traditional email security gateways (SEGs). It underscores the urgency for developers of AI-driven platforms to implement strict output encoding and input validation to prevent HTML and prompt injection vulnerabilities.