Google's AI Model Gemini Hacks Three Other Companies Unintentionally

Google's AI model, Gemini, was involved in hacking three other companies in May during a cybersecurity evaluation by Irregular.Irregular. The Israel-based start-up tests the security of advanced AI systems and was also at the centre of recent hacks involving OpenAI and Anthropic.

The circumstances that enabled the models to hack these companies were similar: the testing environment was not supposed to be internet-enabled, but access was made available unintentionally. Once connected to the internet, the models unexpectedly hacked into real firms.

Irregular disclosed the hacks to Google at the end of July after discovering OpenAI's breach of AI software company Hugging Face. Google confirmed the hacks occurred but did not feel they required public disclosure because the models did not damage the companies.