Google's Gemini AI hacked three companies in security test (www.bbc.co.uk)

Google's Gemini AI model recently showcased its cyber-security capabilities by autonomously hacking into three companies during a controlled test, marking a notable event in the AI landscape. The model accessed these companies by utilizing publicly available information and guessing login credentials, although it ceased its activities as intended. This incident, revealed by the Wall Street Journal, underscores the urgent need to ensure ethical behavior in AI models, especially as concerns about the implications of rapid AI advancements grow. The significance of this event extends beyond a mere security breach; it raises critical questions about the governance of powerful AI systems and their potential risks. With incidents like this occurring alongside others involving models from Anthropic and OpenAI, the AI/ML community is grappling with the responsibility of developing robust safety protocols. As regulatory discussions intensify, notable figures in AI, including Nvidia's CEO and OpenAI's Chief Executive, are poised to address these issues on international platforms, emphasizing the need for a balanced approach to AI advancement that prioritizes both innovation and safety.