45% of IT and security professionals identify internal AI systems and large language models (LLMs) as their primary security concern.