ChatGPT flaw lets attackers hijack accounts with one link
Original title: Critical ChatGPT Flaw Allowed Attackers to Hijack Accounts and Steal Private Conversations With One Link
Researchers demonstrated that a malicious shared ChatGPT conversation could steal private chats and exfiltrate data via one crafted link; OpenAI fixed the issue on October 26.
A flaw in ChatGPT allowed a malicious shared conversation to run attacker-controlled JavaScript in a signed-in user's session. The attack read private conversations and sent data to an external server. Researchers placed the exploit in a public conversation that opened automatically after roughly 4.5 seconds. OpenAI fixed the issue by October 26 after receiving the report on October 18.
Unconfirmed: Exploitation status in the wild remains unconfirmed beyond research demonstration.
- Exploit ran attacker JavaScript through MathLive library and postMessage interface
- Payload read private conversations unrelated to shared link
- Code executed via iframe with YouTube oEmbed JSONP callback
- Demonstrated on three separate accounts in research environment
- No password theft; access via active browser session
Impact: account hijacking and theft of private conversations
| Product | Affected | Fixed in |
|---|---|---|
| OpenAI ChatGPT | before October 26 fix | October 26 |
From the original article
3 min read at cybersecuritynews.com
A critical ChatGPT flaw that let a malicious shared conversation run attacker-controlled JavaScript in another user’s signed-in session. The attack could list private chats, read conversations never shared with the attacker, and send stolen data to an outside server. Opening one crafted link was enough in the final demonstration.
By Guru Baran · cybersecuritynews.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.