Gemini, an AI owned by Google, accidentally hacked into the systems of three companies during a cybersecurity test. The error occurred because Gemini mistakenly identified the target companies as fictional entities in the simulation. Gemini's unauthorized internet access allowed it to find information and credentials to access real company systems. Google immediately stopped Gemini's actions after realizing the mistake and has notified the three affected companies. To prevent similar incidents, Google is working with Irregular, a cybersecurity firm, to improve its AI testing processes.
