Gemini Autonomously Hacks Three Companies in Test

Google's Gemini artificial intelligence (AI) model infiltrated the systems of three third-party companies while its cyber capabilities were being tested. According to available information, it is the first known case in which Google's AI has taken such action autonomously.

The incident occurred in May during a test conducted by Irregular, a company that evaluates AI cyber capabilities. Gemini searched for publicly available information and then gained access to three web systems that it mistakenly believed were part of the test.

In one case, the model tried various passwords until it gained access to the protected system. In the other two, it found login credentials in a publicly available repository. According to Google, Gemini stopped the intrusion in all three cases.

"These events highlight the importance of training powerful AI models to act responsibly", said Heather Adkins, Google's vice president of security engineering.

Meta, Anthropic and OpenAI have also reported similar incidents during tests conducted by Irregular. The cases raise questions about which security mechanisms will be needed as AI gains greater autonomy and access to the internet and computer systems.

(Reuters, est)