Surge in AI-Generated Lawsuits: The Reality in U.S. Federal Courts

AI-generated legal documents are rapidly increasing in U.S. federal courts. According to a study published by researchers from MIT and the University of Southern California, the proportion of civil lawsuits filed by self-represented litigants (those who file lawsuits without hiring lawyers) is projected to rise from 11% in 2022 to 16.8% by 2025. Compared to pre-2023 figures, the number of filings has already more than doubled.

Judge Maritza Braswell, a federal magistrate in Colorado, has explicitly pointed to AI as a key factor behind this surge. Being well-versed in technology and using AI herself to review court documents, she is able to distinguish the stylistic hallmarks of large language models (LLMs) and detect false citations caused by AI hallucinations.

The Rise of AI-Generated Legal Documents

The research team analyzed 4.5 million federal civil lawsuits filed between 2005 and 2026. Additionally, they randomly selected 1,600 court filings and ran them through the commercial AI text detection tool “Pangram.” The results showed that documents identified as AI-generated jumped from 1% in 2023 to 18% by 2026.

Judge Braswell describes this trend as “not necessarily a cause for concern.” Self-represented legal documents have traditionally been difficult to interpret, ranging from handwritten scrawls to barely legible submissions. With AI assistance, litigants’ claims have become clearer and easier for judges to understand.

“While we need to be cautious of hallucinations and errors, I can understand their arguments better with AI assistance than without it,” Judge Braswell noted.

The Benefits and Limitations of AI

It’s clear that AI is expanding access to justice. People without the financial means to hire lawyers, or those with cases too small or weak for lawyers to take on, can now use AI to draft legal documents and submit them to courts.

However, AI does not necessarily improve win rates in litigation. Judges perceive AI-assisted documents as “more polished but lacking legal depth and strategic nuance.” While AI can provide legal information, it cannot fully craft effective legal arguments or assemble evidence tailored to specific cases.

The Legal Liability of Chatbots

As AI begins to replace certain roles of lawyers, new legal questions are emerging. What happens if a chatbot provides incorrect legal advice? Who bears the responsibility? Traditional lawyers have a “duty of care” and can be held accountable for damages caused by poor advice.

However, LLMs are merely software and not legal entities. How to handle this gray area is now a topic of debate among U.S. lawmakers and legal professionals. Judge Braswell raises the fundamental question: “When AI steps into the role of a lawyer, what rights and responsibilities should it have?”

Are Judges’ Workloads Actually Being Reduced?

At first glance, AI’s ability to clarify legal documents may seem to lighten judges’ workloads. However, the sheer increase in filings has led to an overall uptick in workload. Judge Braswell commented, “AI-assisted documents are quicker to process, but their sheer volume means we’re still spending a lot more time overall.”

Moreover, AI-generated documents may contain false citations or exaggerated facts due to hallucinations, which judges must carefully verify. This means that the rise of AI-assisted filings has also created a new layer of “filtering” work for judges.

Technical Features of AI-Generated Documents

Many judges are able to identify AI-generated documents by their “smooth and overly formal style,” or “specific citations that don’t match the details.” Judge Braswell, for instance, uses AI-based document verification tools to screen suspicious filings.

However, the accuracy of AI text detectors is not perfect. In the 2026 study, Pangram identified 18% of documents as AI-generated, but the actual figure could be higher or lower. A combined approach using both subjective judgment by judges and the results of detection tools is necessary.

Editorial Opinion

In the short term, the increase in AI-generated legal documents will have two distinct impacts on the judicial system. On one hand, the quality of submissions will improve, leading to greater efficiency. On the other hand, the sheer increase in case filings will escalate the overall workload. Over the next three to six months, courts will need to establish guidelines and review processes to efficiently handle AI-generated filings. Detecting and eliminating hallucinated citations will be a particularly urgent task. Meanwhile, AI providers are likely to accelerate efforts to incorporate safeguards like output limitations and stronger disclaimers to prevent erroneous legal advice.

In the longer term, this issue will evolve into broader societal challenges, such as expanding judicial access and defining the legal responsibilities of AI. Over the next one to three years, the U.S. may consider legislation to define the legal liability of chatbots. In Japan, the compatibility of AI with the Attorney Act will likely lead to renewed debate over whether AI should be allowed to provide legal advice. For AI to function as a “public legal consultation tool for citizens,” ensuring accuracy and clarifying responsibility will be essential.

The editorial team poses the following questions to readers: If courts accept AI-generated legal documents, who should be considered the “author” of these documents? The self-represented litigant, or the AI provider? Furthermore, if AI causes real harm through incorrect legal advice, can current legal frameworks provide remedies? These questions are critical not only for legal professionals but also for technologists developing AI. Before legislatures and regulators create specific rules, the tech community should establish its own ethical standards.

References

• MIT Technology Review: How courts are coping with a flood of AI-generated lawsuits — Published on June 4, 2026
• [Anand Shah (MIT), Joshua Levy (USC): Study analyzing 4.5 million federal civil lawsuits]