Topic analysis

Critical Vulnerability in ChatGPT for Google Sheets Allows Data Exfiltration and Phishing Attacks

A critical vulnerability in OpenAI's recently launched ChatGPT for Google Sheets extension enables threat actors to conduct indirect prompt injection attacks that exfiltrate user workbooks, deploy phishing overlays, and modify spreadsheets, even when users have disabled automatic edits and enabled human approval requirements. OpenAI mitigated the issue by removing the model's ability to generate Apps Script code following public disclosure from security firm PromptArmor, which did not receive a substantive response to multiple responsible disclosure attempts.