In brief
- LA Superior Court is testing Learned Hand’s AI to help judges prep cases without replacing judicial decisions.
- The company’s CEO warns that AI-assisted legal filings will flood courts with a “bots versus bots” dynamic if left unchecked.
- The system uses a closed set of legal materials and verification layers designed to catch hallucinations before a judge sees the output.
Courts around the world are straining under growing caseloads, and a pilot program in Los Angeles is hoping to change that by testing whether AI can assist judges without offloading their judgment.
The Los Angeles Superior Court is testing an AI tool called Learned Hand that summarizes filings, organizes evidence, and generates draft rulings in civil cases.
The goal is to reduce time spent on administrative tasks so judges can focus on the parts of a case that require legal analysis and discretion, Learned Hand founder and CEO Shlomo Klapper told Decrypt.
“We’re at a place in society where courts are under tremendous strain,” Klapper said. “Their caseloads go up, but no help is coming,” he said, adding that advances in artificial intelligence are “massively dropping the cost of litigation.”
AI is increasing pressure on the courts by making it easier to produce filings, with filings rising 49% from 4,100 to 6,400 in the past year, according to a February 2026 report by national law firm Fisher Phillips.
The Los Angeles Superior Court pilot gives a small group of judicial officers access to Learned Hand’s AI system to test its performance across a case, from intake to draft rulings.
A former judicial law clerk for the U.S. Court of Appeals and deployment strategist with Palantir, Klapper said Learned Hand, founded in 2024, and named after a federal judge of the same name, was designed to give overburdened courts “purpose-built” AI tools that cut down on “drudge work” by surfacing key facts and legal issues while leaving judgment and agency with the human judge.
“With this partnership, we are carefully evaluating emerging technologies to determine how they may support judicial officers in working more efficiently and effectively,” Presiding Judge Sergio C. Tapia II said in a statement. “Let me be clear—while this tool may enhance the way judicial officers review and engage with case files and information, it will not replace, or in any way compromise, the sanctity, independence, and impartiality of judicial decision-making.”
Klapper said the harder part of developing an AI for courts is not generating text but checking AI output against the underlying case materials and legal sources.
“Most of the expense of our large language model is in the verification, not the generation,” Klapper said. “Generation is easy. Anyone can generate something, but how do you make sure that it’s really reliable?”
AI hallucinations have already surfaced in high-profile court cases.
In 2023, the defense team for Prakazrel “Pras” Michel, a founding member of hip-hop group the Fugees, alleged that an AI helped write a closing argument that included frivolous claims and missed weaknesses in the government’s case against him.
That same year, a federal judge ordered lawyers representing former Trump attorney Michael Cohen to provide printed copies of cited cases after the court could not verify them.
Klapper said Learned Hand is built around a narrower pool of source material to reduce the risk of AI hallucinations. Rather than pulling from the open internet or random datasets, the system operates within a defined set of legal materials.
The reason is that large language models can reflect biases in their training data, pointing to examples of AI echoing advice from platforms like Reddit, Klapper said. Learned Hand addresses that by breaking tasks into steps and assigning each step to a model with a specific function.
Learned Hand is also designed so that judges do not need technical training to use it.
“It’s point and click,” Klapper said. “They don’t have to do any prompts.”
Klapper argued that much of a judge’s day is spent on routine tasks rather than legal reasoning, and that the AI aims to allow them to “spend more time on judge work and less time on drudge work.”
Klapper said judges should not take AI outputs at face value and that both the tools and the companies behind them need to prove their reliability.
“I like to say, don’t trust, verify,” he said. “They shouldn’t trust anything. It has to show its worth.”
Daily Debrief Newsletter
Start every day with the top news stories right now, plus original features, a podcast, videos and more.
