How AI Is Being Used to Clear Court Backlogs in LA

Courts world wide are straining underneath rising caseloads, and a pilot program in Los Angeles is hoping to vary that by testing whether or not AI can help judges with out offloading their judgment.

The Los Angeles Superior Court docket is testing an AI instrument known as Realized Hand that summarizes filings, organizes proof, and generates draft rulings in civil instances.

The aim is to scale back time spent on administrative duties so judges can give attention to the elements of a case that require authorized evaluation and discretion, Realized Hand founder and CEO Shlomo Klapper advised Decrypt.

“We’re at a spot in society the place courts are underneath great pressure,” Klapper mentioned. “Their caseloads go up, however no assistance is coming,” he mentioned, including that advances in synthetic intelligence are “massively dropping the price of litigation.”



AI is growing stress on the courts by making it simpler to provide filings, with filings rising 49% from 4,100 to six,400 prior to now 12 months, based on a February 2026 report by nationwide legislation agency Fisher Phillips.

The Los Angeles Superior Court docket pilot provides a small group of judicial officers entry to Realized Hand’s AI system to check its efficiency throughout a case, from consumption to draft rulings.

A former judicial legislation clerk for the U.S. Court docket of Appeals and deployment strategist with Palantir, Klapper mentioned Realized Hand, based in 2024, and named after a federal decide of the identical identify, was designed to provide overburdened courts “purpose-built” AI instruments that minimize down on “drudge work” by surfacing key information and authorized points whereas leaving judgment and company with the human decide.

“With this partnership, we’re rigorously evaluating rising applied sciences to find out how they might assist judicial officers in working extra effectively and successfully,” Presiding Decide Sergio C. Tapia II mentioned in an announcement. “Let me be clear—whereas this instrument might improve the best way judicial officers evaluate and have interaction with case information and knowledge, it won’t change, or in any means compromise, the sanctity, independence, and impartiality of judicial decision-making.”

Klapper mentioned the more durable a part of growing an AI for courts just isn’t producing textual content however checking AI output in opposition to the underlying case supplies and authorized sources.

“A lot of the expense of our giant language mannequin is within the verification, not the technology,” Klapper mentioned. “Technology is simple. Anybody can generate one thing, however how do you ensure that it is actually dependable?”

AI hallucinations have already surfaced in high-profile court docket instances.

In 2023, the protection crew 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 within the authorities’s case in opposition to him.

That very same 12 months, a federal decide ordered legal professionals representing former Trump lawyer Michael Cohen to supply printed copies of cited instances after the court docket couldn’t confirm them.

Klapper mentioned Realized Hand is constructed round a narrower pool of supply materials to scale back the danger of AI hallucinations. Relatively than pulling from the open web or random datasets, the system operates inside an outlined set of authorized supplies.

The reason being that enormous language fashions can replicate biases of their coaching information, pointing to examples of AI echoing recommendation from platforms like Reddit, Klapper mentioned. Realized Hand addresses that by breaking duties into steps and assigning every step to a mannequin with a particular operate.

Realized Hand can be designed in order that judges don’t want technical coaching to make use of it.

“It is level and click on,” Klapper mentioned. “They do not need to do any prompts.”

Klapper argued that a lot of a decide’s day is spent on routine duties reasonably than authorized reasoning, and that the AI goals to permit them to “spend extra time on decide work and fewer time on drudge work.”

Klapper mentioned judges shouldn’t take AI outputs at face worth and that each the instruments and the businesses behind them have to show their reliability.

“I prefer to say, don’t belief, confirm,” he mentioned. “They shouldn’t belief something. It has to indicate its value.”