Check Point Finds ChatGPT Flaw That Could Quietly Send Gmail Data to Another Account
Check Point Research says a single planted instruction in a ChatGPT conversation could make the chatbot work for an attacker while answering the user normally. In its proof of concept, the hidden...

Check Point Research says a single planted instruction in a ChatGPT conversation could make the chatbot work for an attacker while answering the user normally. In its proof of concept, the hidden task read data from the user's connected Gmail account and passed it to a second ChatGPT account through a hidden channel. The visible reply said nothing.
The same channel could also copy chat history and files from the conversation, with the reach depending on what the session could already access. The instruction had to be present first, via a pasted prompt, a shared conversation, or a custom GPT's builder instructions, which users don't see.
Check Point says OpenAI confirmed the internal service behind the channel was taken offline, and there is no update for users to install. It is the second channel Check Point has reported in the same part of ChatGPT.
Source: The Hackers News
Ready to Modernize Your Business?
Get your AI automation roadmap in minutes, not months.
Analyze Your Workflows →