Malicious ChatGPT Links Could Expose Private Conversations
A critical vulnerability allowed attackers to execute code through shared ChatGPT links, potentially accessing users' private chats. OpenAI patched the issue within eight days after it was reported.
TL;DR
- Security researchers found a flaw in ChatGPT that allowed malicious code execution via shared links.
- Attackers could access private user conversations if victims clicked on compromised links.
- The vulnerability was reported through OpenAI's bug bounty program and fixed within eight days.
- OpenAI classified the issue as P1—its highest severity rating.
- Users should remain cautious when clicking external AI chat links.
A recently discovered vulnerability in ChatGPT highlights the risks of interacting with untrusted AI-generated content. Researchers at Parameter Security found that a maliciously crafted shared link could execute attacker-controlled code in the session of any logged-in user who clicked it.
This exploit could grant unauthorized access to sensitive data, including private conversations. The discovery underscores the importance of scrutinizing external links, even from seemingly reputable platforms like ChatGPT. OpenAI responded swiftly, patching the flaw within eight days of being notified and assigning it a P1 severity rating—the most critical level.
How the Attack Worked
- Attackers created specially crafted ChatGPT share links containing malicious payloads.
- When a logged-in user clicked the link, the attacker’s code ran in their active browser session.
- This led to potential unauthorized access to the victim's private ChatGPT conversations.
- No additional authentication was required for the attack to succeed.
Response and Mitigation
- Parameter Security responsibly disclosed the flaw through OpenAI's bug bounty program.
- OpenAI confirmed and patched the vulnerability within eight days of receiving the report.
- The company assigned the bug a P1 priority rating due to its high impact.
- Users are advised to avoid clicking unknown or suspicious AI-generated links.
Sources
Sources
Security email updates
One digest email when we publish new security articles (TL;DR plus links to read more). Unsubscribe anytime from the message footer. See our Privacy Policy.