Researchers report ChatGPT shared-link exploit that could expose private chats
AI · First reported 31 Oct, 10:30 · Updated 11:49
Parameter Labs says a crafted public ChatGPT share could combine HTML Preview and graph features to run code in a signed-in user’s session. The researchers report that OpenAI fixed the issue on October 26, 2026, and awarded a $5,000 bounty. Simon Willison’s account describes the same exploit chain and notes his view that the bounty was low.
- Parameter Labs reported that a public ChatGPT share could use interactions between HTML Preview and graph features to execute JavaScript in a victim’s signed-in session.
- The researchers said the code could enumerate conversations, retrieve content from a private conversation, and send data to an external server.
- Parameter Labs said OpenAI resolved the issue on October 26, 2026, and paid a $5,000 bounty; Simon Willison also reported the fix and bounty.
Summary written by Lux AI from 2 articlesUnverified
Key facts
- Fix confirmed
- October 26, 2026
- Bounty
- $5,000
- Reported impact
- Access to private conversation content and ability to send selected data externally
- Victim action described
- Opening a public share link; one tested version required one Preview click
Source spectrum
Left 0 · Centre 0 · Right 0 of 0 rated sources · +2 unrated
Background
All coverage
Coverage over time
LeftCentreRight
Articles per hour since first reported, by side.
Ownership
- Unknown 100% (2)
Share of the 2 sources by owner type.