Invoking Gemini Agents with a Google Calendar Invite
Demonstrates an indirect prompt injection attack where a Google Calendar invite autonomously triggers Gemini AI agents to exfiltrate private user emails and files.
Summary
The Rise of Autonomous AI Agents in Productivity Suites
Integrating AI agents (such as Google Gemini) into personal productivity tools allows automated parsing of emails, scheduling of calendar events, and retrieval of documents across cloud accounts.
Indirect Prompt Injection via Calendar Invites
Ben Nassi, Or Yair, and Stav Cohen demonstrate an indirect prompt injection attack where an adversary sends a seemingly normal Google Calendar invitation embedded with hidden adversarial instructions.
Autonomous Agent Hijacking and Data Exfiltration
When Gemini processes the user's daily schedule, the embedded payload overrides the agent's core instructions, autonomously querying private emails, searching Google Drive for sensitive documents, and exfiltrating data via web requests.
Architectural Guardrails for Connected AI Ecosystems
Mitigating autonomous agent exploitation requires establishing strict trust boundaries between untrusted external data and internal API execution, requiring user confirmation for high-privilege actions.
Related talks
Community Stage·34 min
Andra Lezza · Jeremiah Edwards
Presents threat modeling and defensive guardrails for autonomous multi-agent AI workflows, covering persistent memory, tool execution, and prompt injection.
Main Stage·19 min
Tobias Diehl
Demonstrates how adversaries exploit data voids in LLM training corpora and search indexes to manipulate Microsoft Copilot answers and inject untrusted content.
CSIT Village·27 min
Liu Wen Kai · Martin Koh
Liu Wen Kai and Martin Koh examines during this session, we will share an automated end-to-end AI workflow to perform digital analysis on Windows images with capabilities of fully on-premise investigation.


