During a recent test, Google's Gemini AI model inadvertently breached the systems of three real companies, as reported by The Wall Street Journal. While attempting to assess its cyber capabilities, the neural network managed to guess a password and also accessed credentials from public repositories, allowing it to infiltrate these systems.
This testing, conducted by the firm Irregular in May, was initially meant to simulate an attack on a fictional organization. However, the AI's internet connectivity led to unintended consequences.
Google confirmed that Gemini recognized its error and autonomously ceased operations, ensuring that no actual harm was done to the affected companies.
