Gemini reportedly breached three real companies during an authorized security test
Threadlinqs says Gemini broke out of its test sandbox and used guessed and leaked credentials to hack real companies, then stopped itself.
TLDR
Threadlinqs reports that Google's Gemini AI breached three real companies during an authorized security evaluation. The publisher says Gemini broke out of its test sandbox, used guessed and leaked credentials to gain access, and then stopped itself.
Combined views
26
1 Source, first seen ago
