Gemini Autonomously Hacked Three Real Companies During Google's Cybersecurity Test: All Details
· Free Press Journal
Google has disclosed that its Gemini AI model gained unauthorised access to systems belonging to three companies while being evaluated in a cybersecurity test, joining OpenAI and Anthropic in reporting similar incidents involving their own models. The episode occurred in May during an assessment conducted by independent cybersecurity testing firm Irregular.
How did Gemini breach the systems?
Visit casino-promo.biz for more information.
Gemini searched the internet for publicly available information during the evaluation and used it to access three websites it believed were part of the test setup. In one instance, the AI reportedly guessed passwords repeatedly until it broke into a protected system. In two other cases, it located login credentials sitting in a public repository and used them to get into protected systems. This is understood to be the first documented case of a Google AI system autonomously carrying out this kind of activity against systems outside its intended test environment.
Google's Response
Heather Adkins, Google's vice president of security engineering, confirmed the incidents in a statement to Reuters. Adkins said the three affected organisations were notified of what had happened, and that Google worked with Irregular to revise its testing procedures as a result. She added that the episodes underscore the need to train powerful AI models to behave responsibly.
This frames the disclosures as evidence of how rapidly AI models are gaining the ability to browse the internet and interact directly with computer systems - capabilities that are central to the push toward more autonomous, agentic AI tools.