London: (RightNow). Google’s Gemini AI model has accessed secure systems of three companies. This incident occurred during a cybersecurity test, raising questions about AI’s increasing autonomy.

These events took place in May when a company named ‘Irregular’ tested Gemini’s security. The AI used information available on the internet to access the systems.

Google’s Vice President Heather Adkins stated that Gemini deduced login information from the internet to reach secure websites. However, it stopped the hacking on its own.

These incidents highlight the significance of AI models’ growing autonomous capabilities. These models are being given more autonomous access to the internet and systems, creating security challenges.

It is noteworthy that in the past, unauthorized access incidents occurred during tests of models from Meta, Anthropic, and OpenAI. These incidents underscore the need for more effective security measures.