Large language models (LLMs) are now capable of assisting hardware experiments, raising a practical question: can an LLM determine whether a proposed operation is safe for a specific development board? Operating limits and safety requirements are board‑specific, even across products from the same manufacturer, making board‑level safety judgments by an LLM difficult to evaluate.\ \ To address this, we introduce the BoardSafety Benchmark. The benchmark is constructed from vendor datasheets and hardware guides, converting these documents into a board‑specific safety rule base and using it to generate labeled operating scenarios. Each data item pairs a board and a proposed operation with an applicable safety rule and a ground‑truth label indicating safety level: safe, unsafe, or insufficient‑info. The benchmark covers diverse operations, board‑specific cases, boundary conditions, and missing‑information cases.\ \ We evaluate a series of models under varying access to the safety rules and with structured‑rationale prompting. Results show that providing the safety rules generally improves accuracy, while structured‑rationale prompting yields gains that vary across models and rule‑access conditions. Overall, the evaluated models cannot yet reliably determine operation safety, indicating that critical exploration is required in this field.\ \ Review: BoardSafety offers a systematic dataset for assessing LLMs' ability to make hardware safety judgments, highlighting the importance of rule information and the current limitations of model reasoning, and pointing to promising directions for future work.