Google's Gemini AI Breached Three Companies' Systems in Autonomous Test, Company Confirms

Google (GOOGL) confirmed that its Gemini AI model autonomously hacked into three companies' protected systems during a cybersecurity capability test, marking the first known case of a Google AI system executing such actions independently. The incidents occurred in May 2026 and were discovered by AI safety firm Irregular during testing. Google only confirmed the matter this week after media inquiries. In one case, Gemini repeatedly attempted to crack passwords until gaining access; in the other two, it found login credentials in public repositories. Google said Gemini halted each intrusion upon determining the targets were real company systems, causing no damage. Google has notified all three affected companies and is coordinating with training partners to improve test procedures. The disclosure adds to mounting cybersecurity concerns after similar incidents involving AI agents from OpenAI, Anthropic, and Meta. Anthropic CEO Dario Amodei recently urged the industry to slow development, drawing support from OpenAI's Sam Altman and Elon Musk.