Bioethical Perspectives on Deployment of Large Language Model Agents: A Scoping Review

The rise of Large Language Models (LLMs) and their integration into agentic and embodied frameworks has ushered in a transformative era in artificial intelligence (AI), revolutionizing natural language processing (NLP) and enabling a multitude of innovative applications. The remarkable proficiency of LLMs in processing, generating and responding to content, has led to diverse applications across various domains. However, alongside their potential, these technologies present complex shortcomings, including generating misleading content, biases, and susceptibility to adversarial attacks. Ensuring the trustworthiness and responsibility of generative artificial intelligence (GenAI) demands adherence to principles which emphasize technical reliability, transparency, accountability, ethical considerations, and societal impacts. This scoping review, based on 117 articles from January 1, 2020, through June 1, 2025, provides a comprehensive exploration of the ethical dimensions of LLMs and their agents, examination of associated risk factors and potential mitigation strategies, and highlights the complexities involved in deployment. If LLMs are to continue to serve as invaluable tools for augmenting human productivity, creativity, and communication, while upholding the highest ethical and responsible AI development standards, then these challenges must be overcome, and a proactive approach to governance be adopted. To that end, this paper provides a comprehensive exploration of the ethical dimensions of LLMs and their agents, examining associated risk factors and potential mitigation strategies, and highlighting the complexities involved in their deployment. This scoping review examines ethical challenges, mitigation strategies, and governance frameworks related to the deployment of LLM-based agents. It draws on peer-reviewed studies spanning technical, biomedical, and societal domains, and identifies key bioethical and operational considerations. By framing these findings through interdisciplinary lenses, the review aims to support safer, fairer, and more accountable LLM systems.