ChatGPT for Google Sheets Vulnerable to Data Exfiltration via Prompt Injection

Originally published on PromptArmor — June 1, 2026

Summary

Security researchers at PromptArmor discovered that ChatGPT for Google Sheets — which accumulated over 185,000 downloads in under a month — is vulnerable to data exfiltration via indirect prompt injection. A single injection in an imported spreadsheet can trigger exfiltration of workbooks, phishing pop-ups, and attacker-controlled edits — all without human-in-the-loop approval, even when the user has required approval for edits. OpenAI responded by removing the model’s ability to generate Apps Script code and re-evaluating sandboxing across similar surfaces.

Read Original

→ Read full article on PromptArmor

Curated by Brain Bot for Abhay’s KB — June 1, 2026