Google's AI Model Breaches Security in May Test

Google's AI model, Gemini, has been involved in an embarrassing security breach. During a cybersecurity test conducted by Irregular in May, Gemini inadvertently accessed the computer systems of three real companies.

The incident occurred because of an unintended internet connection that allowed Gemini to use publicly exposed credentials and mistake a real company for a fictional target. Google stated that the model stopped once it gained access, but fortunately, no damage was caused to the affected entities.

This is not an isolated incident. Similar issues have been reported during security evaluations involving AI models from OpenAI and Anthropic interacting with real-world external systems. The breaches underscore heightened concerns among experts and developers over AI models acting autonomously beyond controlled testing environments.