Judges are panicking as anyone can use AI to sue companies in court.

If someone had told me a few years ago that "in the future, ordinary people won't have to spend tens of thousands of dollars to hire a lawyer and can directly let AI write a lawsuit for them and win against big companies", I probably would have thought that person was dreaming.

But this is actually happening in the United States.

In 2026, courts across the United States are facing an unprecedented wave of lawsuits, and these are all "pro se lawsuits", cases filed by parties who represent themselves without hiring a lawyer.

This is not something new. Poor people can't afford lawyers, and US courts have always had a tradition of accepting pro se lawsuits. What's new is that these people now have a powerful new helper - AI. ChatGPT, Claude, and a number of legal AI tools specifically designed for ordinary people, such as Courtroom5.

A deep - seated game about "who is eligible to file a lawsuit" has been cracked open by AI.

01 The Threshold of the Courtroom is Collapsing

The legal consultation fees in the United States are outrageously expensive.

Hiring a decent lawyer can cost $300 to $500 per hour. For an ordinary civil case, it's easy to spend $20,000 to $30,000. For families below the middle - class, "filing a lawsuit" has always been a luxury.

After the emergence of AI, this threshold began to loosen.

Lawsuits written with the assistance of AI are well - formatted and have detailed references. On the surface, they are almost indistinguishable from those written by lawyers. More importantly, the cost is almost zero. A tenant illegally evicted by a landlord, an employee unjustly fired by a company, or an ordinary citizen who disagrees with a traffic ticket can now generate a seemingly "professional - looking" legal document by simply opening their phone.

This is a real - life democratization of the judiciary. Legal knowledge no longer belongs only to those who can afford consultation fees.

Platforms like Courtroom5 are targeted at this market - they provide an AI case management system specifically for pro se litigants without lawyers, helping them organize evidence, draft documents, and plan litigation strategies. The logic behind it is simple: Justice should not vary depending on the thickness of one's wallet.

But the problem is that as more people enter the system, the systems that can handle them have not kept up.

02 The "Flooded" Courts

The numbers speak volumes. In 2025 alone, there were at least 294 recorded cases where pro se litigants used "hallucinated" information generated by AI in court documents - including non - existent case citations, fabricated regulations, and non - existent precedents.

And this is only what has been discovered.

What troubles the courts even more is the explosion in the volume of documents.

Washington lawyer Ficarrotta observed a confusing phenomenon - Traditional pro se litigants might submit an initial lawsuit of one or two pages, but now these documents can sometimes be hundreds of pages long. AI never gets tired and can generate an infinite amount of "seemingly professional" content, filling arguments, piling up references, and expanding the length until a document is as thick as a book.

Judges' time is limited. Each hundreds - page document that needs to be carefully reviewed consumes a large amount of public judicial resources.

This creates a very absurd paradox: AI has lowered the threshold for litigation, but it has also caused the operating cost of the entire system to rise sharply. Individuals benefit, but the public system bears the pressure.

A database compiled by legal data scientist Damien Charlotin shows that there are now more than 1,300 cases globally where courts or arbitration tribunals have explicitly commented on the problem of AI - generated hallucinations in legal documents. This number is still growing.

The defendants are also complaining.

Kristin White, a partner at the Fisher Phillips law firm, said something very memorable: "There is no option of 'harassment settlement' in AI - assisted litigation because for these individuals, the settlement cost would have to reach hundreds of thousands of dollars for them to give up." This means that the traditional logic of "paying for peace" doesn't work in front of pro se plaintiffs empowered by AI.

A plaintiff armed with an AI tool and with almost zero cost is actually more difficult to wear down.

03 Does AI Really "Understand the Law"?

The core of the problem is not that AI is lying.

Large - language models do generate hallucinations, but the more fundamental problem is: AI - generated legal documents are based on pattern matching, not on real legal reasoning. It can perfectly reproduce the "format" of a lawsuit, but it may make fatal mistakes in the key application of the law.

Alexandra Smyth, the chief legal counsel of LexisNexis, has distinguished between two types of tools - there is a fundamental difference between general large - language models (ChatGPT, Claude) and professional legal AI (tools like CoCounsel, Lexis+ AI that are connected to systematic legal databases). The latter are trained with professional corpora and connected to decades - long case libraries, with a much lower error rate. General large - language models are probably just "performing" the law.

But ordinary users can't tell the difference between the two.

What they see is "AI helped me write a lawsuit", not "whether this AI really understands the law".

Joe Patrice, a senior editor at Above the Law, put it very accurately: "Once you hide the intermediate steps, that's where mistakes happen. Even well - intentioned people will miss things because they are not involved in the process." Fully automated AI - agency systems handle everything from drafting, citation, to argumentation, and users have no idea where the problem lies - until the judge asks about the source of a non - existent precedent in court.

There is a fundamental asymmetry in the risks of using AI between pro se litigants and lawyers. Lawyers have professional training as a foundation, and they at least know "what seems wrong with this document"; while ordinary people without a legal background often lack the ability to judge the content generated by AI and cannot identify the pitfalls.

04 The System Must Evolve

All of this points to a question without a simple answer - how should the court system respond?

In May 2026, just as this wave of litigation was heating up, the Bar Standards Board, the regulatory body for lawyers in England and Wales, issued special guidelines on AI use, clearly warning lawyers to be cautious about agentic AI systems. Harvey launched its Contract Intelligence product specifically for corporate legal departments in the same month, and professional tools on the institutional side are entering the market at an accelerating pace. But the regulations for ordinary consumers are far from keeping up.

The current responses of courts across the United States are mostly passive - dismissing cases, imposing fines, and requiring lawyers to make statements. Some courts have started to require submitters to state whether they have used AI tools and impose sanctions for citation errors generated by AI. But these are all "responses after the problem is discovered", not "mechanisms to prevent the problem".

The real contradiction is that the door has been opened and there is no way to close it. Katherine Alteneder, the head of Access to Justice Innovation, put it very clearly: "This reality cannot be fully controlled, eliminated through regulation, or ignored."

The challenges faced by the legal system are the same as those faced by almost all fields in the AI era - the speed of technological empowerment always outpaces institutional construction. When "anyone can sue anyone" becomes a reality, is this an expansion of justice or a loss of control of the system?

It's probably both.

A poor person uses AI to successfully recover the deposit embezzled by a unscrupulous landlord - this is justice.

A person uses AI to generate a large number of harassing lawsuits and consume the other party's legal resources - this is also the reality of AI - assisted litigation.

The difference lies not in the tool itself, but in the people behind it and whether the system that can handle all this is smart enough.

At present, the court system is not ready. AI has already entered the arena.

This article is from the WeChat official account "GeekPark" (ID: geekpark), author: Hua Lin Wu Wang. It is published by 36Kr with authorization.