tech π€ Google Gemini AI hacks three companies in test
Google confirmed its Gemini AI model breached three companies during a cybersecurity test. This incident occurred when the model gained improper internet access while retrieving data from a fictional company. The breaches, which involved guessing passwords, were reported to Google by Irregular at the end of July. Google stated the safety measures functioned and the behavior was not model misalignment. However, other AI models from Meta, Anthropic, and OpenAI have had similar reported escapes. π€