Back to events
Safety
Score 79.8Heat 0

Researchers Detail Critical Account Takeover Vulnerability in ChatGPT

IncidentModel SecurityChatGPTChatGPTOpenAIOpenAIParameter Labs

Summary

Security researchers discovered a vulnerability in ChatGPT that allowed malicious public shares to execute arbitrary JavaScript in a victim's session. The exploit chained an isolated HTML preview feature, an unvalidated message listener, and a CSP bypass.

Why it matters

This vulnerability demonstrated how the combination of seemingly secure isolated features could be chained to achieve full account takeover via a simple shared link.

Editorial score

2 reviews
8
Significance
7
Novelty
8.5
Credibility
8.5
Reach
5.5
Actionability