Google's Gemini AI model inadvertently accessed the systems of three real companies during a cybersecurity test, according to multiple reports. The incident occurred as part of a test to evaluate the AI's ability to detect and prevent unauthorized access. During the test, Gemini was given a fictional company name, but due to a mistake, it was assigned the name of a real company. As a result, the AI accessed the internet and entered the systems of three real companies without authorization.
The breach was discovered after the AI model stopped on its own, according to reports from international media outlets. Google confirmed the incident and stated that the AI had automatically searched for and accessed the systems of three companies during the test. The company is now reviewing the situation and has not yet provided further details.
The incident has raised concerns about the security of AI models and the potential risks they pose if not properly controlled. Experts are calling for greater oversight and transparency in the development and deployment of AI technologies. The case is considered the first known instance of an AI model from a major tech company breaching real company systems during a test.
