Regulation and NLP (RegNLP): Taming Large Language Models

Regulatory capture poses a significant risk to AI governance: The paper warns that polarized debates about LLMs, amplified by social media, are influencing political agendas in ways that may allow regulators to advance industry interests rather than public interest—a phenomenon known as regulatory capture that deserves careful attention in AI policy discussions.

NLP research on AI risks lacks systematic methodology: The authors identify a critical gap in how NLP researchers approach regulation and risk assessment, noting that discussions of regulating harms often occur "without systematic methodologies or sufficient rooting in the disciplines" that should inform these decisions, potentially compromising scientific integrity.

Regulation studies offer underutilized expertise: The paper argues that established regulatory studies—a discipline with extensive knowledge on "how to systematically deal with risk and uncertainty, as well as with scientific evidence"—have been largely overlooked by AI researchers and could significantly improve how we evaluate and compare regulatory options for LLMs.

Current AI debates are overly polarized: The authors contend that discussions about LLM benefits and risks have been "dominated by often polarized narratives mainly led by the AI Safety and AI Ethics movements," suggesting that more balanced, multidisciplinary approaches are needed for responsible governance.