The Impact of Generative AI on Judicial Systems

The advent and rapid spread of generative artificial intelligence tools, such as ChatGPT and Claude, are redefining various sectors, and the judicial system is no exception. Preliminary research, still in pre-print, suggests that the adoption of these technologies has coincided with a significant increase in pro se legal cases, meaning situations where a defendant or plaintiff represents themselves in court without the aid of an attorney. This phenomenon, while seemingly a step towards greater access to justice, is generating an unprecedented workload for U.S. federal courts.

The authors of the study, Anand Shah and Joshua Levy, in their work titled "Access to Justice in the Age of AI: Evidence from U.S. Federal Courts," highlight how the ease of use of LLMs has lowered the barrier to entry for self-representation. What previously required specific legal expertise and high costs can now be addressed with the help of an AI assistant, capable of generating legal documents, identifying statutes, and guiding through procedures.

A Data and Trend Analysis

The research is based on a large volume of administrative data, including over 4.5 million non-prisoner civil court cases between 2005 and 2026, and 46 million related PACER (Public Access to Court Electronic Records) docket entries. The results show stability in the percentage of pro se cases around 11% until 2022. However, after the widespread diffusion of LLMs like ChatGPT, this percentage began to grow rapidly, reaching 16.8% in 2025.

This surge is not limited to the number of cases. The study also notes a 158% increase in intra-case activity for each pro se case, measured by the volume of entries in the court docket, such as filings and motions. This means that each self-managed case requires significantly more effort from judges and court staff. Furthermore, the increase is primarily driven by plaintiffs, who use AI to file complaints, with annual counts doubling from approximately 19,705 to 39,167 between 2015 and 2025, while defendant-side pro se cases slightly decreased.

Implications for the Judicial System and Data Sovereignty

The authors' argument is that, although the study is descriptive and does not establish a direct causality between the use of specific LLMs and individual cases, the temporal correlation is difficult to ignore. The use of AI detection software, such as Pangram, on a sample of 1,600 complaints showed an increase in AI usage from "essentially zero" in the pre-AI period to over 18% in 2026. This scenario poses a critical challenge for courts, which cannot easily increase their capacity or refuse to hear cases. The current judicial backlog is already a persistent feature of the federal system, and no influx of new judges is anticipated to meet this growing demand.

The issue also raises broader questions about data sovereignty and compliance. Although the article does not directly focus on these aspects, the use of external LLMs for preparing legal documents implies the processing of sensitive information. For organizations and individuals evaluating the adoption of AI solutions, the choice between cloud deployment and self-hosted or on-premise options becomes crucial, especially in contexts requiring strict controls over data privacy and location.

Future Prospects and Trade-offs

The dilemma is clear: on one hand, generative AI offers an unprecedented opportunity to democratize access to justice, allowing more people with legitimate grievances to assert their rights. On the other hand, such a drastic increase in cases and their complexity risks clogging the system, lengthening resolution times for everyone, including non-AI-assisted cases. Joshua Levy, one of the authors, suggested that a possible solution could be to allow judges themselves to use AI tools for "standardized" or "templatable" work, while maintaining human judgment as the central element.

This trade-off between accessibility and efficiency is a recurring theme in the age of artificial intelligence. As technology continues to evolve, it is essential for existing systems to adapt, exploring innovative solutions that can balance the benefits of automation with the need to maintain the integrity and functionality of fundamental institutions. The challenge for the judicial system will be to find a balance that allows it to harness the potential of AI without compromising its ability to deliver justice in a timely and effective manner.