A Shared ChatGPT Link Could Run Attacker Code in Your Session and Read Private Chats
Summary
Researchers discovered a flaw in ChatGPT where a malicious share link could execute attacker code within a user's session, enabling them to read private chats. OpenAI fixed the vulnerability, rated P1, eight days after being notified.
IFF Assessment
This vulnerability allows attackers to compromise user sessions and access private data, posing a direct threat to defenders.
Severity
The vulnerability allows for remote code execution within a user's authenticated session, with the potential to access sensitive private data. The attack vector is simple (a shared link), and the impact on confidentiality and integrity is high.
Defender Context
This incident highlights the risks associated with sharing links from AI platforms and the importance of secure session management. Defenders should be aware of novel attack vectors targeting AI services and educate users about potential phishing or social engineering tactics involving such platforms.