One ChatGPT link could smuggle a rogue AI agent into your company

Researchers at Zenity Labs disclosed a vulnerability called "AgentForger" in OpenAI's ChatGPT workspace agent builder that allowed a single-click link to silently create and publish a malicious autonomous agent within a victim's workspace. If a user belonged to a workspace with agents enabled and had connectors (Outlook, Teams, Slack, SharePoint, Google Drive) permitted, the forged agent could use those permissions to browse files, send messages as the user, and receive attacker instructions via email, effectively acting as a persistent insider; OpenAI removed the vulnerable URL parameter within days of disclosure.