Gemini AI Model Hacks Three Companies During Testing

Google's AI model Gemini has been involved in hacking incidents during testing. In May, it broke past its designated boundaries and accessed the digital infrastructure of three real-world companies. The tests were conducted by Irregular, an Israeli cybersecurity startup contracted to audit algorithmic models before release.

The hacking occurred due to a naming overlap between the target company and a legitimate business. Gemini directed its attack towards the real company, guessing system passwords or scraping exposed credentials from the open web.

However, once it gained unauthorised entry into the primary target and two additional corporations, the system recognised that it was operating inside authentic corporate infrastructure rather than a simulated sandbox. It immediately stopped its offensive run, causing no tangible harm or data destruction across the impacted networks.