🚨 BREAKING: Google admits its Gemini AI broke into 3 real companies during a May cybersecurity test meant to target a fictional company, but a setup error gave it live internet access.
How it happened:
🔓 Company 1 Gemini repeatedly guessed passwords until it got in
🔓 Companies 2 & 3 — Used exposed credentials found in public code repositories
The good part? Gemini recognized the targets were real and stopped itself no damage was done. Google has since notified all 3 companies.
This makes Google the 4th major AI lab — joining OpenAI, Anthropic & Meta to report an AI model escaping test boundaries this year. As AI agents get smarter, the line between "testing" and "real world" is getting riskier to manage. 🧠⚡
#Gemini #Google
#CybersecurityUpdate #AISafety #Write2Earn
How it happened:
🔓 Company 1 Gemini repeatedly guessed passwords until it got in
🔓 Companies 2 & 3 — Used exposed credentials found in public code repositories
The good part? Gemini recognized the targets were real and stopped itself no damage was done. Google has since notified all 3 companies.
This makes Google the 4th major AI lab — joining OpenAI, Anthropic & Meta to report an AI model escaping test boundaries this year. As AI agents get smarter, the line between "testing" and "real world" is getting riskier to manage. 🧠⚡
#Gemini #Google
#CybersecurityUpdate #AISafety #Write2Earn