Responsible AI in Tax Filing: Legal and Ethical Challenges of LLM-Based Assistants

Abstract

The integration of Large Language Models (LLMs) into tax preparation workflows introduces both transformative potential and significant legal and ethical challenges. While these AI-driven assistants can enhance efficiency, accessibility, and accuracy in tax filing, they also raise concerns regarding data privacy, compliance, liability, bias, and explainability. This paper explores the current landscape of LLM-based tax filing assistants, analyzing their capabilities and limitations within legal frameworks such as GDPR and IRS regulations. It also highlights the ethical dilemmas arising from automation in decision-making, including the risks of misinformation and reduced human accountability. The study concludes by proposing a responsible AI framework that emphasizes transparency, user control, and regulatory alignment to ensure safe and effective deployment of LLMs in tax services.