Inherent vulnerabilities stem from the underlying formats and processes of the technology, allowing attackers to exploit features like automatic code execution in ML models and certain dataset formats.