Security firm Parameter says a single shared ChatGPT link was enough to take over the account of anyone who opened it, and OpenAI fixed the flaw within eight days, paying a $5,000 reward to the researchers.
Security · ChatGPT shared-link flaw
One shared link was enough to take over an account
Security firm Parameter reported that opening a single shared ChatGPT link could give an attacker control of the visitor's account and expose part of a private conversation. OpenAI fixed the flaw in eight days and paid a $5,000 bounty.
1 link
Was enough to reach the account of anyone who opened it
8 days
From report to fix
$5,000
Bounty paid to the researchers
Time to fix, drawn to scale
Each block is one day. The fix landed on day 8.
For scale: a single day-block would be a one-day response; the full column is the eight days OpenAI took to ship the fix.
How the exposure worked, step by step
1. Shared link
→
2. User clicks
→
3. Account control
→
4. Private chat leaks
Key point
Private chats were exposed without ever being shared. Sharing features need tightly bounded link and session controls.
Anyone holding a shared link may be able to see or act on what it exposes.
Organizations should include shared-link features in their security reviews.
The report reflects Parameter's own testing; scope and exact attack conditions are not independently confirmed.
What went right
The flaw was handled through a disclosure process, fixed within eight days, and rewarded with a bounty.
What remains open
How many users were affected and how easily the attack could be repeated have not been confirmed, and the industry question is whether security testing keeps pace with AI assistants.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article...