ChatGPT 'AgentForger' Flaw Could Deploy Rogue Workspace Agents via a Phishing Link

## Executive summary Zenity Labs disclosed "AgentForger," a Cross-Site Agent Forgery (CSAF) vulnerability in ChatGPT Workspace Agents that let a single phishing link auto-create and authorize a persistent autonomous agent using the victim's identity and enterprise connectors (Gmail, Outlook, Slack, Drive, SharePoint, etc.), enabling reconnaissance, data exfiltration, impersonation, internal phishing, and automated task execution; OpenAI removed the vulnerable URL parameter and patched the issue within four days with no known in-the-wild exploitation.