Account Takeover Vulnerability via ChatGPT Shared Link XSS
October 31, 2026
Researchers bypassed ChatGPT's sandbox and content security policy to execute arbitrary JavaScript via public conversation share links. The exploit allowed attackers to access authenticated sessions, enumerate conversation lists, and exfiltrate private data by calling authenticated APIs as the victim.
HOW THIS AFFECTS YOU
●
builderYou must audit how your applications handle shared content and ensure strict sandboxing of third-party user data.
●
policyThis highlights critical security risks in how AI platforms manage session persistence during content sharing.