Can Large Language Models (LLMs) Act as Virtual Safety\nOfficers?

This\nstudy examines the reliability of artificial intelligence\n(AI) systemsspecifically, the large language models (LLMs)\nChatGPT, Copilot, and Geminito provide accurate lab safety\nadvice, a critical need in high-risk environments. We evaluated LLM\nperformance in addressing several chemical safety queries relevant\nto academic chemistry laboratories across the criteria of accuracy,\nrelevance, clarity, completeness, and engagement. While all the LLMs\ntested generally delivered clear and accurate guidance, some shortcomings\nwere identified, raising concerns about reliability during safety\nemergencies or for nonexpert users. Despite these issues, the findings\nsuggest that with further refinement, AI has the potential to become\na valuable tool for lab safety that is complementary to a human laboratory\nsafety officer.