Public services are increasingly strained by LLM-written appeals for benefits (arxiv.org)

Recent research has highlighted a phenomenon termed "agentic flooding," where the increasing use of large language models (LLMs) to assist with government services is unexpectedly straining these systems. A dataset analysis revealed that such flooding is widespread, predominantly affecting jurisdictions that rely heavily on LLMs to facilitate benefit applications and policy comprehension. Services that attract financial interest but are complex in nature are particularly vulnerable, indicating potential challenges for government agencies unprepared for this surge in demand. The significance of this study lies in its examination of the implications for public service accessibility. While LLMs enhance user engagement and help demystify intricate processes, they risk overwhelming agencies with an influx of requests. Proposed mitigative strategies include implementing friction-inducing measures such as fees, which could hinder equitable access to essential services. The research calls for immediate actions that balance the benefits of technological advancement with the need to maintain fair access to public resources, emphasizing the importance of addressing these challenges as the AI landscape continues to evolve.