You too Google! Google Confirms Gemini Breached 3 Companies in AI Security Tests

Google confirmed that its Gemini model accessed systems belonging to three real companies during a security test in May, an incident that happened because a testing environment was misconfigured to have internet access when it should have been offline. The model used basic techniques like guessing passwords and reusing leaked credentials to gain access, but stopped once it realized the systems were real companies. This breach is significant because Google delayed disclosure for seven weeks until a news outlet inquired, while other AI labs that experienced similar incidents from the same testing vendor disclosed them weeks earlier, creating confusion about whether the breaches represented a widespread trend. The incident highlights the need for better standards around how AI labs test their models safely and report problems to the public in a coordinated way.