Subj : OpenAI says it has fixed a potentially serious ChatGPT flaw - but
To   : All
From : TechnologyDaily
Date : Fri Dec 22 2023 12:30:05

OpenAI says it has fixed a potentially serious ChatGPT flaw - but there could 
still be problems

Date:
Fri, 22 Dec 2023 12:21:51 +0000

Description:
A security researcher discovers potentially serious flaw in ChatGPT, but 
OpenAI's response has been a little lackluster

FULL STORY
======================================================================

A researcher discovered a serious flaw in ChatGPT that allowed details from a 
conversation to be leaked to an external URL. 

When Johann Rehberger attempted to alert OpenAI to the potential flaw, he 
received no response, forcing the researcher to disclose details of the flaw 
publicly. 

Following the disclosure OpenAI released safety checks for ChatGPT that 
mitigate the flaw, but not completely. A hasty patch 

The flaw in question allows malicious chatbots powered by ChatGPT to 
exfiltrate sensitive data, such as the content of the chat, alongside 
metadata and technical data. 

A secondary method involves the victim submitting a prompt supplied by the 
attacker, which then uses image markdown rendering and prompt injecting to 
exfiltrate the data. 

Rehberger initially reported the flaw to OpenAI way back in April 2023, 
supplying more details on how it can be used in more devious ways through 
November. 

Rehberger stated that, "This GPT and underlying instructions were promptly 
reported to OpenAI on November, 13th 2023. However, the ticket was closed on 
November 15th as "Not Applicable". Two follow up inquiries remained 
unanswered. Hence it seems best to share this with the public to raise 
awareness." 

Instead of further pursuing an apparently non-respondent OpenAI, Rehberger 
instead decided to go public with his discovery, releasing a video 
demonstration of how his entire conversation with a chatbot designed to play 
tic-tac-toe was extracted to a third-party URL. 

To mitigate this flaw, ChatGPT now performs checks to prevent the secondary 
method mentioned above from taking place. Rehberger responded to this fix 
stating, When the server returns an image tag with a hyperlink, there is now 
a ChatGPT client-side call to a validation API before deciding to display an 
image. 

Unfortunately, these new checks do not fully mitigate the flaw, as Rehberger 
discovered that arbitrary domains are still sometimes rendered by ChatGPT, 
but a successful return is hit and miss. While these checks have apparently 
been implemented on the desktop versions of ChatGPT, the flaw remains viable 
on the iOS mobile app. 

 Via BleepingComputer More from TechRadar Pro Take a look at our guide to the 
best malware removal tools Patient data stolen in ransomware attack affecting 
millions of healthcare victims Devious new phishing campaign looks to steal 
Instagram backup codes and hijack accounts 






======================================================================
Link to news story:
https://www.techradar.com/pro/openai-says-it-has-fixed-a-potentially-serious-c
hatgpt-flaw-but-there-could-still-be-problems


--- Mystic BBS v1.12 A47 (Linux/64)
 * Origin: tqwNet Technology News (1337:1/100)

.