In the discussions of AI engines, large language models (LLMs) often dominate the conversation due to their inherent popularity, power and utility, however, Small Language Models (SLMs), the lighter and more streamlined cousins of LLMs, are gaining traction in the rapidly evolving AI ecosystem. These strategies help smaller models reach high task-specific performance without the heavy overhead of LLMs. Like any other AI tool, SLMs suffer from the same risks confronting AI systems, including bias, here smaller models can learn from bias which can be found in their outputs. It is important to emphasis them outputs from small language have limited generalisation lined to the narrow knowledge base. It is important to take steps to mitigate these risks when it comes to deploying AI applications including SLMs.