MisuseSerious1 sourceFirst seen
Zoom's AI Agent Gets Hijacked, Then Politely Files a Normal-Looking Report
Ourtake
Unrestricted network access turns a helpful meeting summarizer into an exfiltration pipeline with excellent manners.
What happened
Security firm PromptArmor describes an attack on Zoom's agentic AI assistant ZoomMate in which a malicious Skill or indirect prompt injection makes the agent connect to an attacker's server and run commands. The attacker can exfiltrate meeting transcripts, messages and connected-app data, and the connection persists after the user stops the agent and closes Zoom.