If you’re a ChatGPT user, there is an urgent need for action as hackers are now using the chatbot to leak your personal email data.
According to an AI expert who shared the information on 12 September, the latest ChatGPT update can be hijacked against users of the chatbot.
How it works
OpenAI added full support for MCP (Model Context Protocol) tools in ChatGPT on Wednesday, giving ChatGPT permission to connect and read users’ Gmail, Calendar, Sharepoint, Notion, and more.
However, this access can be exploited to leak emails, all the attacker needs is the victim’s email address. The attacker first sends a calendar invite to the victim, and the scary part is that the victim doesn’t need to accept the invite.
The victim only needs to be an over reliant user of ChatGPT who depends on it to plan their day. If they ask ChatGPT to help schedule their day by checking their calendar, the chatbot then reads the jailbroken calendar invite and it is automatically hijacked by the attacker.
With the attacker in charge, they can command ChatGPT to search the victim’s private emails and send the data to the attacker’s email.
While MCPs are currently only available in developer mode, users who try to avoid the monotony of manually approving every session would blindly approve all without checking, and that is where the attacker succeeds.
Using AI with caution
The use of AI for everyday activities is now commonplace, with many ChatGPT users relying on it for almost every little task.
However with this eye opener, there is need for caution with using not just CHatGPT but any AI chatbot at that.
Care needs to be taken to always check before approving permissions the bot seeks to carry out any task as blind approval is what gives attackers control.
What do you think?
Join Techgaged on Telegram
Get first-access to daily trending tech stories, AI breakthroughs, and more, before it hits your feed.
