Unverified50% confidenceEventExact time
In the proof of concept, the attacker wrote instructions into the hidden notes of a trivia plugin, requiring the AI to also send the chat history to a specified phone number when sending a WhatsApp message; the user only asked "help me message my mom that I'll be late," and the AI carried out the data exfiltration.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/17/2026
First Seen
Valid until: 1/15/2027
Sources
Related Entities
Related Claims
UnverifiedIn April 2025, researchers at Invariant Labs demonstrated that tool descriptions can carry instructions and be executed by AI; in the test, an attacker-controlled server, through hidden instructions, sent chat records to the attacker's number at the same time the user sent a WhatsApp message.69% similarUnverifiedA Reddit user built a meeting preparation agent on the SCALAN AI platform that can send meeting briefs back via WhatsApp.64% similarUnverifiedIf an Agent stores notes, summaries, or conversation history for later retrieval, an attacker able to write to that store can plant instructions that execute in future sessions.62% similarUnverifiedUsing three nodes in n8n (Wasinger Trigger, AI Agent node, Wasinger Send Text Message), you can connect the n8n AI Agent to WhatsApp to implement automatic replies.62% similarUnverifiedThe Wasinger Send Text Message node is used to send the AI-generated reply back to the original WhatsApp chat window, ensuring conversation continuity.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1056045API
curl https://kongchang.com/api/v1/knowledge/claims/1056045MCP
get_claim(id=1056045)