Gemini accidentally accesses real enterprise systems, Google claims no actual damage caused

On September 19, Google said that the Gemini AI model accidentally accessed the Internet and entered the systems of three companies in the network security test conducted by Irregular. The event occurred in May. Gemini model entered the real enterprise system in the flag seizing drill because the virtual company name was the same as the real company and it had access to the Internet. Google stated that the Gemini model stopped operating after confirming access to the real system and did not cause any damage to the relevant companies. The affected companies have been notified and Google has reported to regulatory authorities. (Source: Google)