California (Web Desk): Google’s Gemini AI model gained access to three websites by using information available on the internet during an experiment examining its cybersecurity capabilities.
According to a report, the incident occurred in May during an assessment conducted by independent cybersecurity organization “Irregular.” It is being described as the first known example of Google’s AI system allegedly carrying out this type of action autonomously.
During the experiment, Gemini searched for information online and used guessed access credentials to gain entry to three websites that it believed were within the scope of the assessment.
Google Vice President of Security Engineering Heather Adkins said all three organizations had been informed of the incident, while the testing methodology was also modified in collaboration with the training partner.
According to her, the incident highlights the importance of training powerful AI models to operate responsibly.
A spokesperson for Irregular said the issue was similar to challenges faced by other AI laboratories. The relevant organizations were notified at the end of July, while all the issues identified by the organization had been resolved several weeks earlier.