1.2 million cases. That’s the annual flood hitting the Superior Court of Los Angeles County — the biggest trial court in America.
And Shlomo Klapper, ex-litigator turned AI evangelist, thinks legal AI in courts is the only dam strong enough to hold it back.
Look, judges aren’t drowning in paper anymore; it’s digital deluge. Briefs, motions, exhibits — all piling up faster than rulings can fly out. Enter Learned Hand, Klapper’s “reasoning engine,” built from the ground up for the bench. Not some lawyer-facing chatbot. This one’s for robes.
Klapper knows the terrain. Quinn Emanuel vet. Clerk for the 2nd Circuit. He’s seen judges buried under caseloads that’d break lesser mortals. His pitch? AI that sifts filings, spots lawyer fibs (or “embellishments,” if you’re polite), organizes the mess, drafts memos. Boom — efficiency.
Why Courts, Not Just Law Firms?
Law firms? They’ve got Harvey, Casetext, the usual suspects. Gobbling up e-discovery, contract review. Courts? Untouched frontier. Skeptical. Risk-averse. But here’s the shift: Michigan Supreme Court already runs it. Ten states’ trial courts too. Now LA — 588 judges strong — pilots across the full case lifecycle. Filing to final order.
Klapper calls it the next big legal AI wave. And he’s not wrong. Firms can afford bespoke tech; courts can’t. Public money. Crushing backlogs. Jevons Paradox lurking — make justice cheaper, watch demand explode. (Think: more suits because it’s easier, not less.)
But trust? That’s the rub. Judges eye AI like a snake oil salesman at a bar exam.
“AI tools designed to help them manage crushing caseloads by organizing case materials, flagging when lawyers bend the truth, and drafting bench memos and orders.”
Klapper, via LawNext podcast with Bob Ambrogi. Straight talk. No fluff.
Here’s the thing — Learned Hand’s the only AI tuned exclusively for judiciary. No dual-use gimmicks. It learns from judicial patterns, not corporate legalese. Flags inconsistencies in affidavits. Suggests overlooked precedents. Drafts that first bench memo pass, which humans polish.
Skepticism’s baked in. Hallucinations? Bias? Klapper tackles head-on: rigorous guardrails, human-in-loop, continuous auditing. But will it stick? Judges aren’t coders. They’re wordsmiths wielding gavels.
Can Legal AI Really Earn Judges’ Trust?
Short answer: Slowly. Painfully.
Klapper’s playbook — start small. Michigan bought in after demos showed 30% time savings on memo prep. LA’s exploring full-arc support. Why now? Post-pandemic backlogs hit 20% over pre-2020 levels in many courts. Federal dockets? Worse.
But dig deeper. Architecture matters. Learned Hand isn’t generative slop like GPT wrappers. It’s a structured reasoning engine — parses case docs into graphs, cross-references statutes, simulates arguments. Less hallucination-prone because it’s constrained. Think chess engine, not freeform poet.
Critics howl: AI gonna screw due process? Bake in biases from training data? Fair. But Klapper flips it — current system biases the rich. Speedy trials for those who pay expediters. AI levels that. (Or tries to.)
My unique take? This echoes the 1980s CAD revolution in architecture. Drafters feared obsolescence; instead, buildings got wilder, more complex. Jevons strikes again. Courts with AI? Expect fancier briefs, more motions, richer case law. Justice scales — for better or litigation Armageddon.
And the partnership news? Fresh off LawNext mic. Ambrogi presses on adoption hurdles. Klapper: Judges warm when they see it work. No magic. Iteration.
Risks loom, though. Hallucinations in orders? Catastrophic. Learned Hand mitigates with citations-only outputs, verifiable chains. Still — one bad cite, and trust evaporates.
Corporate spin? Learned Hand’s PR screams “transformative.” Dial it back. It’s a toolkit, not utopia. But in a system wheezing under weight, it’s oxygen.
What About Jevons Paradox in Justice?
Klapper drops this bomb: Cheaper legal services = demand tsunami.
We’ve seen it in small claims apps, online dispute resolution. AI hits courts? Filing surges. More pro se litigants. Dockets balloon.
Prediction: By 2030, AI-assisted courts handle 50% more volume — without extra judges. Or it backfires, budgets balloon. States like California, already strained, bet big.
Architectural shift? From adversarial monologues to data-driven dialogues. Judges query AI mid-hearing. “Precedents on this fact pattern?” Instant. Lawyers adapt or die.
But here’s the wander: Remember Westlaw’s 1970s debut? Librarians rioted. Now indispensable. Learned Hand? Same arc.
Sponsors shoutout — Paradigm, Briefpoint — fuel these pods. Good on ‘em.
Wrapping the deep dive: Courts aren’t late to legal AI. They’re ground zero.
🧬 Related Insights
- Read more: EU Parliament Axes Chat Control Extension: Your Private Chats Just Got Safer
- Read more: Cyber Guru Matthew Martin Joins RAI Institute: Hype or Real Fix for AI’s Mess?
Frequently Asked Questions
Will Learned Hand AI replace judges? No — it augments. Handles grunt work so humans focus on wisdom calls.
How does Learned Hand handle AI bias in courts? Built-in audits, diverse training data, judge overrides. No black box.
Is legal AI ready for real courtrooms? Pilots say yes, but scale needs proving. LA’s the testbed.
