OpenAI's Upcoming Astra AI Model May Possess 'Critical' Cybersecurity Capabilities, Company Warns

OpenAI has disclosed that its forthcoming AI model, Astra, may exhibit 'critical' cybersecurity capabilities, including the autonomous identification and exploitation of zero-day vulnerabilities and complex cyberattacks. This assessment, based on internal evaluations and expert reviews, has prompted OpenAI to pause some internal development and implement stricter safety protocols, including isolated testing environments and enhanced monitoring.