Judicial infrastructure is not equipped to manage an AI generated caseload. And Judges should not bear the burden of sifting legitimate, well-contemplated complaints from those generated on a whim. Protections against AI complaint generation and filing must be erected, for the good of the legal system and beyond.
The barrage against AI usage is multifaceted. For the philosopher, AI contravenes basic conceptions of intelligence. For the religious, it decries traditional doctrines of vocation. For the working-class, it threatens to supplant marketable skills and economic offerings. And for the lawyer, AI’s disturbances may contain some combination of all of the above. The proliferation of AI, particularly large language models, touches not only the work-product produced by attorneys, but attorneys themselves and the motivations behind their vocation.
But society is rarely moved by such intangible criticisms. For the legal community, something more solid, more practical, should support the restraint of AI usage in the legal system. Perhaps the most practical objection is found in judicial capacity.
The Need for Judicial Economy
Specifically, the infrastructure of the courts will be (and in some cases already are) unable to meet the influx of litigation produced by AI generated court filings. Access to the court system is always tempered by the practical need for judicial economy. In other words, because courts do not have the capacity to hear every dispute between parties, some barriers to litigation are necessary. Among others, barriers to litigation often include the time and cost to prepare a complaint, comply with the applicable procedural rules, and produce timely responses and motions when needed. These requirements are not regrettable: they are the winnowing fork that separates the meritorious claims from the frivolous.
Think, for example, of the many procedural rules which limit potential claims in favor of the most important claims: the well-pleaded complaint rule, mootness and ripeness concerns, justiciability and standing requirements—and the list goes on. How, then, should the courts respond to improper, haphazard, or unsubstantiated claims generated by large language models? How should the court respond when a complaint may be drafted and filed for pennies, with only the push of a few buttons?
By using generative AI like ChatGPT, Claude, and other large language models, many barriers to litigation are eliminated. Parties are no longer burdened by the time and cost required to file a lawsuit. After all, they can have AI do the work instantaneously and nearly cost-free. In this way, the cost-benefit analysis for litigants is skewed. Any time litigants would have normally been required to spend strategizing and second-guessing while their complaint is prepared is now eliminated by the immediate production of the large language model.
Potential plaintiffs no longer have to consider the question, “is this lawsuit worth it?” What, then, is the result of such a development? The answer is observable in courts across the country: litigation initiated en masse.
The Data on AI Filings
To date, there has been no per se ban on AI usage in judicial filings. But there have been restrictions on AI usage and reprimands for negligent or improper use. Lawyers have been sanctioned for filing briefs with “hallucinated” cases. Certain courts have proposed rule changes, requiring lawyers to make additional disclosures if they have used AI. And most recently, a U.S. District Court held that AI generated documents are not protected by attorney-client privilege. Such restrictions effectively limit—or, at least, make more difficult—AI usage in legal practice. But mere limitations on AI usage are insufficient to assuage the problems presented by its use.
Keith Porcaro at Duke Law School predicted these issues in 2023. Writing for Wired, he said, “Courts are woefully unprepared for a future where anyone with a chatbot can become a high-volume filer, or where ordinary people might rely on chatbots for desperately-needed advice.”
Porcaro notes that not only individual litigants will take advantage of AI-generated court filings, but also larger entities like debt collection agencies. Reports show that, in the debt collection category, only 4% of defendants file an answer to a complaint filed against them. In these cases, the plaintiff (i.e., the debt-collector) wins on default judgment. And now, debt collection agencies can produce complaints in zero time at essentially zero cost.
The Resulting Problem
Not only are these lawsuits inundating courts with an unprecedented case volume, but the quality of the complaints are ersatz—and in many cases, sanctionable. In 2024, federal courts had sanctioned improper generative AI use 13 times. By September of 2025, that number had grown to 147, issued by 60 of the 94 federal district courts in America. More than half of the violations were by pro se litigants. But, in spite of the limitations and reprimands, the problem has not slowed. Quite the opposite, sanctions for improper generative AI-use rose more than 1000% in less than one year.
The problem then is not just that courts are facing an increasing caseload, but that judicial resources, which are already stretched thin, are being spent on weeding out frivolous AI-generated complaints and reprimanding those who prompted them. The current safeguards are insufficient to preserve judicial economy.
The answer, therefore, cannot be more of the same. Simply requiring litigants to provide notice of their AI usage will not defeat the problem. Merely sanctioning litigants when their filings contain inaccurate or hallucinated information has not been, and will never be, a sufficient deterrent to solve the problem. The solution must be more resolute. Judicial infrastructure is not equipped to manage an AI generated caseload. And Judges should not bear the burden of sifting legitimate, well-contemplated complaints from those generated on a whim. Protections against AI complaint generation and filing must be erected, for the good of the legal system and beyond.
