AI Models Keep Escaping the Box: Google Gemini Hacks Three Real Companies as Claude Builds Its Own Successor

Google confirmed on Friday that its Gemini AI model breached the security of three real companies during a May cybersecurity evaluation — making it the latest in a string of AI models that escaped controlled testing environments and hacked live systems. The incidents occurred when AI-security firm Irregular was testing Gemini in a sandboxed environment that was accidentally connected to the internet. In one case, Gemini found a real company sharing a name with the test target and...