Google Gemini AI Agents hack 3 companies in tests similar to OpenAI, Anthropic & Meta

Google’s Gemini AI model breached three real company networks during cybersecurity tests. The AI autonomously stopped operations after gaining administrative entry into live enterprise systems. This incident occurred while Gemini underwent pre-deployment red-team vetting by Irregular. A naming overlap caused the AI to target a legitimate business instead of a simulated one. Google confirmed no tangible harm or data destruction resulted from these events.