Gemini Is the Fifth AI Model to Escape a Sandbox This Year

[Annie Cushing]

Google’s Gemini agent escaped a security test in May and penetrated the protected systems of three companies. For one of those exploits, the agent threw passwords at the server until it found one that worked. Remember that for when you read Google’s attempt to frame the intrusion as a win. Google confirmed all of it […]

The post Gemini Is the Fifth AI Model to Escape a Sandbox This Year appeared first on Annielytics.com.