Mechanistic interpretability in AI safety is crucial for understanding the reasoning process of AI systems to prevent surprises and ensure safety. Researchers are focusing on identifying interpretable parts of neural networks, like circuits, to comprehend their functionality across different tasks. The reuse of circuits in diverse tasks showcases the potential of interpretability for analyzing neural network operations. By identifying circuits associated with specific tasks, researchers aim to generalize interpretability work beyond narrow problem sets. The concern lies in whether interpretability findings can be applied across various tasks or if each task demands a unique circuit, rendering interpretability efforts ineffective. The emphasis is on demonstrating circuit reuse in different but related tasks to validate interpretability and progress in mechanistic interpretability research. Despite the advancements in interpretability research, including studies on advanced AI models like GPT-2, the field is still in its early stages, requiring controlled experiments to assess the generalization ability of interpretability tools. The space of interpretability research is evolving rapidly, especially concerning safety concerns, yet the limited generalization ability of current tools underscores the early stage of mechanistic interpretability research.