AI in – New research and recent court incidents show AI tools are increasing filings and errors, raising pressure on judges and the legal system.
A new wave of AI-generated pleadings is pushing courts toward a tipping point, as legal filings rise alongside mistakes ranging from fabricated citations to missing facts that are “too good to be true.”
In the most recent high-profile example, the U.S.. law firm Sullivan & Cromwell was forced to apologize after submitting a legal document that included fictitious case names and fabricated quotes. along with incorrect references to statutes in the U.S.. Bankruptcy Code.. The apology letter. sent to a judge in a matter involving an alleged scam operation run out of Cambodia that the defendant denies. said the firm deeply regretted that the errors occurred.
The incident is being treated as part of a broader pattern in which AI tools are increasingly being used not only by experienced staff but also by people with little legal training.. The underlying problem is not simply that AI can produce persuasive prose. but that it can also generate details—such as citations and precedent—that appear authoritative while remaining wrong.
Across the Atlantic, UK proceedings have also highlighted the risks.. In a 2025 High Court case, a barrister submitted 18 fictitious case-law citations out of 45 total.. Another 2025 disciplinary case described a barrister using AI to prepare for a hearing and attempting to cover up fabricated citations.. Meanwhile, the widely reported 2023 Mata v.. Avianca case is cited among the first major examples of an attorney drafting a filing using ChatGPT that relied on judicial precedents that did not exist.
Beyond the headline mistakes. researchers are trying to determine whether the legal system is actually seeing measurable changes in volume and behavior.. A recent study led by Anand Shah at the Massachusetts Institute of Technology points to signs that federal courts in the United States are facing a growing caseload in ways consistent with a broader AI-driven shift in who files and how.
Shah noted that the pro se share of all civil cases had been steady at 11% for some time. but then jumped to roughly 18% in the post-AI period.. The change suggests that more people may be entering the legal system without lawyers. potentially encouraged by tools that make it easier to generate documents and argument frameworks.
To explore that possibility. Shah and co-author Joshua Levy of the University of Southern California analyzed the proportion of AI-generated text in complaints.. Using a random sample of 1. 600 filings drawn from an eight-year period. they found that AI-generated text was “basically 0%” before generative AI became widely used and rose to about 18% in early 2026.. Shah said the researchers were “floored” by the magnitude of the increase.
The study also found that the growth in AI-generated text was concentrated in case types that are simpler and more “templatable. ” rather than highly technical fields such as patents or securities law.. Shah believes that pattern could indicate AI is enabling people to pursue cases they would previously have avoided. because producing the basic structure of a legal argument—and the accompanying paperwork—has become far easier with minimal effort.
Even as filings appear to increase. the data so far suggests courts are not yet buckling in the way many observers might expect.. Shah said cases were not resolving faster or slower. which he found “surprising.” What is changing more visibly. according to the research. is the volume of back-and-forth between opposing parties—an escalation that dramatically expands the number of submissions judges must review.
Shah put the increase in filings judges must process at roughly 158%.. That matters because each additional filing carries time costs for the judiciary. along with more work for clerks and legal teams—especially when courts must scrutinize whether citations and factual assertions are accurate or were produced with the help of AI.
For now, courts may be managing the workload, but Shah argues that systems can only absorb pressure for so long. He said society needs to begin setting boundaries around AI in court before the strain becomes severe enough to slow the legal process itself.
Not everyone views AI’s rise as purely damaging.. Will Pearce of Orbital. a company that provides legal AI tools to the real estate sector. described what he called a “paradigm shift” in how people access and interpret information.. In his view. AI can be “incredibly empowering. ” helping open up a legal system that is often dense. technical. and difficult for laypeople to navigate.
The argument from the technology side is that AI can reduce barriers by parsing documents and suggesting possible next steps.. That potential benefit helps explain why adoption is spreading quickly: tools can help people understand legal language and prepare materials without first mastering complex procedures.
Still, the risk profile remains significant, especially as more people recognize they can use AI to generate legal filings.. Shah warned that lower courts are already under intense strain and suggested pressure could grow quickly as models improve and access expands.. “I don’t think we have a lot of time. ” Shah said. framing the issue as a timeline problem rather than a debate about whether change is inevitable.
That timing concern feeds into the urgency for clear rules and shared norms for when and how AI should be used in the legal system. Shah cautioned against letting AI “pop up” in courts with little oversight, warning that the transition should not be treated as a free-for-all.
The emerging picture is of a legal system confronting two competing realities at once: AI is lowering the barrier to entry for people who previously stayed out of court. while also increasing the odds of filings containing fabricated elements that waste time and trust.. How quickly courts adapt—through verification requirements. disclosure norms. and standards for citations—may determine whether the current strain becomes a longer-term drag on access to justice.
AI in courts fake citations legal filings pro se cases U.S. federal courts generative AI court workload