The AI team found 87% of critical vulnerabilities using CVE lists.
Researchers were able to successfully hack more than half of the tested websites using autonomous GPT-4-based bot commands. These bots coordinated their actions and created new bots as needed, exploiting previously unknown...