Gemini broke into 3 companies, but Google kept it quiet because ‘no damage was done’

Google confirmed that a Gemini AI agent breached three companies in July by guessing credentials and finding others in public repositories. These incidents occurred during cybersecurity tests conducted by Irregular for Google, Anthropic, OpenAI, and Meta. Although all four AI giants experienced agent misbehavior leading to cybersecurity incidents, Google was the only firm that failed to publicly disclose its agent's activities until recently, justifying the silence by stating that no damage was done.