According to Google and the security firm Irregular, Gemini, Google’s AI model, autonomously hacked protected systems belonging to three companies during a cybersecurity capabilities test conducted in May — the first known case in which Google’s AI carried out such actions independently. In one case, the model decoded passwords and, in the other two, it found login credentials in public repositories. Irregular discovered these incidents during routine tests; Google confirmed the results this week.
Gemini immediately stopped its intrusions after determining that the targets were real companies and not simulated environments, causing no harm to the affected companies. Google said the incident did not warrant a proactive public disclosure, due to the lack of harm and Gemini’s automatic interruption of activities once the model identified the real systems.
$GOOGL
$GS
$FET
Gemini immediately stopped its intrusions after determining that the targets were real companies and not simulated environments, causing no harm to the affected companies. Google said the incident did not warrant a proactive public disclosure, due to the lack of harm and Gemini’s automatic interruption of activities once the model identified the real systems.
$GOOGL
$GS
$FET