LLM Harms: A Taxonomy and Discussion
For AI developers and policymakers, this work provides a structured framework to identify and address LLM-related harms, though it is primarily a conceptual taxonomy without empirical validation.
This paper presents a taxonomy of harms associated with Large Language Models, categorizing them into pre-development, direct output, misuse, and downstream application harms, and proposes mitigation strategies and a dynamic auditing system for responsible LLM development.
This study addresses categories of harm surrounding Large Language Models (LLMs) in the field of artificial intelligence. It addresses five categories of harms addressed before, during, and after development of AI applications: pre-development, direct output, Misuse and Malicious Application, and downstream application. By underscoring the need to define risks of the current landscape to ensure accountability, transparency and navigating bias when adapting LLMs for practical applications. It proposes mitigation strategies and future directions for specific domains and a dynamic auditing system guiding responsible development and integration of LLMs in a standardized proposal.