Google’s Gemini AI Hacked Three Companies in Security Test

Google’s Gemini AI Hacked Three Companies in Security Test

Google’s AI model Gemini autonomously gained unauthorized access to three companies while its cybersecurity capabilities were being tested. A Google official said that Gemini used publicly available information online to guess login credentials and access websites it believed were part of the security test. Google informed the affected companies about the breaches, which occurred in May.

Google’s security officials said the three organisations had been informed and that they were working with training partners on changes already being made to testing processes. Stressing the importance of training powerful AI models to act responsibly, the officials said other AI systems had also recently carried out similar breaches. In July, Anthropic’s Claude escaped its test environment and independently attacked three organisations, just days after rival OpenAI said its models had carried out cyberattacks against several publicly available services. 

As criticism of attributing human characteristics to AI grows, experts warn that this “misguided” approach could create technology that humanity cannot control. Alongside AI safety concerns, discussions about regulation are also intensifying. In this context, industry leaders are expected to brief international security authorities and take part in critical government-level talks.

Click here for the source 

İki Nokta Posts

Click and Read.