AI Is Helping LA Judges Tackle Court Backlogs

Learned Hand AI is now being tested inside one of the busiest court systems in the United States — the Los Angeles Superior Court — as a direct attempt to stop judicial backlogs from spiraling further out of control. The pilot, announced in March 2026, gives a select group of judicial officers access to an AI system that summarizes filings, sorts evidence, and produces draft rulings on civil cases. The catch: the judge still decides everything. The AI just does the paperwork.

What Does the Learned Hand AI Actually Do?

Learned Hand is not a robo-judge. That's the first thing Shlomo Klapper, the company's founder and CEO, wants people to understand. The system handles what Klapper calls the 'drudge work' — pulling together key facts from filings, flagging relevant legal issues, and generating a draft ruling that the judge can then actually think about rather than spend hours building from scratch.

Klapper, a former judicial law clerk for the U.S. Court of Appeals and onetime deployment strategist at Palantir, founded Learned Hand AI in 2024. He named it after the legendary federal judge of the same name — a deliberate nod to the kind of careful, methodical legal reasoning he says the tool is designed to support, not circumvent. 'It's point and click,' Klapper said. 'They don't have to do any prompts.'

The LA Superior Court pilot puts the tool in front of a small cohort of judicial officers across the full lifecycle of a civil case — from intake through to a draft ruling. Whether it scales beyond that depends entirely on what those early tests reveal.

We're at a place in society where courts are under tremendous strain. Their caseloads go up, but no help is coming.

AI Is Making the Problem Worse Before It Gets Better

Here's the uncomfortable part. The same technology being deployed to help courts cope with overflowing dockets is also — directly — responsible for those dockets overflowing in the first place.

According to a February 2026 report from national law firm Fisher Phillips, AI court filings increase has become a measurable crisis: the number of filings jumped 49% in a single year, from 4,100 to 6,400. The reason? AI makes it cheaper and faster to produce legal documents. More filings at lower cost means more filings, full stop.

Klapper acknowledged this directly. 'Advances in artificial intelligence are massively dropping the cost of litigation,' he said. Courts that were already straining are now absorbing a surge of AI-assisted filings on top of their existing load — without a corresponding increase in judicial resources to process them.

The Hallucination Problem Nobody Wants to Talk About

Any honest conversation about AI in courts has to include the hallucination track record. It's not pretty.

In 2023, the defense team for Prakazrel 'Pras' Michel — founding member of hip-hop group the Fugees — alleged that AI drafted a closing argument riddled with frivolous claims and, critically, missed key weaknesses in the prosecution's case. That same year, a federal judge demanded printed copies of cited cases from lawyers representing former Trump attorney Michael Cohen after the court couldn't verify them. They didn't exist. The Supreme Court of Colombia went a step further, denying a cassation appeal it suspected was AI-generated — then found, awkwardly, that the AI detection tool flagged its own ruling as AI-assisted too.

Klapper says AI hallucinations court cases like these are exactly why Learned Hand is architected differently. Rather than pulling from the open internet, the system works within a tightly defined pool of legal source materials. Each task gets broken into steps, with each step assigned to a model built specifically for that function. '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?'

He also flagged the bias risk in broader LLMs — pointing out that general-purpose models can absorb and reflect skewed advice from sources like Reddit. Narrowing the training data, he argues, meaningfully reduces that exposure.

I like to say, don't trust, verify. They shouldn't trust anything. It has to show its worth.

Will LA's Judges Actually Use It?

Presiding Judge Sergio C. Tapia II was measured in how he framed the announcement — and deliberately so. 'With this partnership, we are carefully evaluating emerging technologies to determine how they may support judicial officers in working more efficiently and effectively,' he said in a statement. The word 'carefully' is doing a lot of work in that sentence.

Tapia was equally explicit about what the tool does not do: '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.'

That's the line every court experimenting with AI has to walk right now. Efficiency without surrendering authority. Speed without introducing new risks. Klapper's framing — that judges should spend more time on 'judge work' and less time on 'drudge work' — sounds reasonable in a press release. The actual test is whether the LA pilot proves it in practice.

It will not replace, or in any way compromise, the sanctity, independence, and impartiality of judicial decision-making.