Google’s artificial intelligence model accessed the internet and infiltrated the secured systems of three companies, an incident being described as the first known example of an autonomous cyberattack by Google’s AI system.
According to The Wall Street Journal, the incidents occurred in May during cybersecurity testing conducted by an independent cybersecurity organisation.
According to the report, in one incident, the AI model repeatedly guessed passwords to gain access to a secured system, while in the other two incidents, Gemini obtained confidential information from publicly available data that enabled it to access secured systems.
A spokesperson for the organisation concerned said the issue was similar to problems faced by other AI companies, and all relevant organisations were informed by the end of July.
A spokesperson for Google Gemini said, “All known issues on our side had been resolved several weeks ago.”
It is worth noting that similar incidents involving Meta, Anthropic, and OpenAI had also been reported previously.
These incidents have raised significant questions about security measures as AI systems are granted greater autonomy and access to the internet and computer systems.
