AI Assistant Parameter-to-Prompt (P2P) Injection Link via rovoChatPrompt URL Parameter

Ankit Mehta@Secvyn

•updated Aug 11, 2026•3•0•8

This rule detects potentially malicious AI prompt injection attempts targeting web-based AI assistants like Microsoft Copilot or Atlassian Rovo. It monitors CommonSecurityLog data for specific URL parameters (rovoChatPrompt, q, prompt) containing suspicious keywords (e.g., 'ignore', 'system', 'exfiltrate') or unusually long parameter values, which are indicators of efforts to bypass AI safety guardrails or perform data exfiltration via prompt manipulation.