*AI Security Skills to Prioritize by 2026*

The author highlights that the rapid integration of AI features (such as LLMs and GenAI) into production environments introduces security risks, often stemming from basic issues: prompt injections, excessive permissions for agents, data leaks via connectors, or accidental exposures in RAG systems. They note that these vulnerabilities are more related to application security (appsec) and cloud security than to machine learning (ML) or data science expertise. The discussion focuses on valuable skills, the effectiveness of hands-on labs versus theoretical learning, and adapting threat modeling to address these emerging technologies.