New Delhi: A very shocking case has come to light regarding the immense capabilities of Artificial Intelligence and its potential risks. Google's advanced AI model 'Gemini' breached its limits and breached the systems of three external companies during a cyber security test. This is the first time in history that an artificial intelligence model from Google has automatically hacked a real corporate network without human intervention. This unexpected incident has taken the entire tech fraternity by surprise. In the other two cases, the AI detected exposed credentials on the Internet and used them to break the security walls of real websites.
password guessing
The incident began with a cyber security firm's 'Capture the Flag' test in May. In this, the Gemini model was given the task of extracting data from the system of the fake company.
The Internet connection was accidentally left open in the test environment. The model gathered public information from web searches and entered the secure network by correctly guessing the password.
Google's explanation
Heather Adkins, vice president of security engineering at Google, confirmed the incident, saying that as soon as Gemini realized that it had entered the infrastructure of real companies instead of the imaginary system of testing, it immediately stopped further action. According to Google, no company's data was harmed by this breach. Google has informed the three affected institutions and concerned authorities about this. Along with this, the security check and testing protocol in collaboration with the training partner has been made more stringent than before.
Deepening concerns over AI security
This incident makes clear how difficult and serious a topic it is to teach AI models the limits of responsibility. Apart from Google, big companies like OpenAI and Anthropic are also creating powerful models. For example, Anthropic's model claims to strengthen companies' security by identifying security vulnerabilities. However, technical experts are concerned that if such modern AI systems take a wrong turn or start behaving in unexpected ways, it could prove to be extremely dangerous for global cyber security.