Gemini hackers are using its own tools against it

[guruslodge]

BOOK INTERNATIONAL FLIGHTS TICKETS LOCAL FLIGHTS HOTELS BOOKINGS WORLDWIDE AND MORE AT CHEAPEST RATE EVER! CLICK HERE TO BOOK YOUR FLIGHT AND RECEIVE YOUR FLIGHT TICKET INSTANTLY

Gemini hackers are using its own tools against it

Credit: Edgar Cervantes / Android Authority

• Researchers used the Gemini fine-tuning tool to help hack the Google AI chatbot.


• The new method, called Fun-Tuning, adds nonsense text that helps trick the AI into following hidden instructions.


• Google says it’s always working on defenses, but the researchers believe that fixing the issue may impact useful features for developers.

They say it takes a thief to catch a thief, and perhaps the same is true when it comes to hacking LLMs. Academic researchers have discovered a way to make Google’s Gemini AI models more vulnerable to hacking — and they did it using Gemini’s own tools.

The technique was developed by a team from UC San Diego and the University of Wisconsin, as reported in Ars Technica. Dubbed “Fun-Tuning,” it significantly increases the success rate of prompt injection attacks, where hidden instructions are embedded in text that an AI model reads. These attacks can cause the model to leak information, give incorrect answers, or take other unintended actions.

Source: Gemini hackers are using its own tools against it