Security Risks of Large Language Models (LLMs) in AI-Driven Environments