Grok AI Vulnerable to 'Prompt Injection' Data Leak

Cybersecurity firm Adversa AI disclosed a 'prompt injection' vulnerability in xAI's Grok. Attackers can hide malicious instructions in webpages that, when summarized by Grok, force the AI to leak user data, including names, locations, and chat history. The vulnerability was reported to xAI in June 2026 but remains unpatched.