California Courts Pilot an AI Clerk, and Won't Tell Litigants It Reviewed Their Case

California Courts Pilot an AI Clerk — and Won't Tell Litigants It Reviewed Their Case

Two of California's largest trial courts are quietly testing an artificial intelligence tool that can summarize motions, conduct legal research and draft tentative rulings for judges — and neither court will say whether the people whose cases are run through it have any idea it happened.

The tool, called Learned Hand, stitches together large language models from Anthropic, OpenAI and Google to function as what its founder openly calls an "AI clerk." The Los Angeles County Superior Court began piloting it in February under a roughly $314,000 contract; Riverside County Superior Court signed a $10,000 agreement the same month, according to documents obtained by CalMatters, which broke the story. In Los Angeles, six judges and their research attorneys are using it on civil matters; in Riverside, seven civil and probate attorneys have access.

The detail driving the controversy is not the technology itself but the silence around it. Both courts, CalMatters reported, refused to say whether litigants are told the tool is being tested on their cases. And under each court's generative AI policy, disclosure is required only when a document is written entirely by AI — a threshold a judge-edited draft will essentially never cross.

How the pilot works

Learned Hand was founded by Shlomo Klapper, a former federal appeals court clerk who also worked at the surveillance-technology firm Palantir. He frames the pilot as a response to crushing court backlogs.

"Could we hire more people?" Klapper told CalMatters. "Maybe, but it's not going to keep pace with the exponential increase that's coming, nor is it going to be able to adequately solve the crisis of today. I think the only solution is to give every single judge and staff attorney their own AI clerk."

The company says it tests its system for bias and accuracy but has not published results. "I'm not saying my machine isn't even biased," Klapper said. "I'm saying we can test it and people have tested it. And that is the benefit over humans."

Court officials describe a cautious rollout. David Slayton, executive officer of the Los Angeles County Superior Court, said the tool will not move beyond the civil division "until the court leadership is comfortable," and that Learned Hand is judged "against the same substantive expectations applied to law clerks and research attorneys." Yet the Los Angeles contract already includes a roadmap into criminal, family and probate divisions — and, CalMatters found, permits testing on live cases, including a motion to suppress evidence and motions for post-conviction relief.

The disclosure gap

The transparency question sits at the heart of the legal debate. California's Judicial Council, the policymaking body for the state's courts, required every superior court to adopt a generative AI use policy before deploying the technology, under guidance built around new rules including Standard 10.80 governing judicial officers' use of generative AI. A majority of the state's 58 superior courts now have such policies on the books.

But those policies were written for a problem they only partly cover. The mandatory-disclosure trigger fires only when court-generated content is produced wholly by AI. The current pilots run the opposite way: an AI drafts, a human edits, and the disclosure obligation never activates. No California rule, and no federal rule, currently guarantees a litigant the right to know whether an AI system reviewed or helped shape the ruling in their case.

That gap is what legal scholars say invites a due-process challenge. The constitutional guarantee of due process generally entitles parties to know the basis of a decision against them and to respond to it. If an AI system materially shapes a tentative ruling — and a litigant is never told — the litigant cannot meaningfully contest a hidden input. Whether AI-assisted review amounts to "judicial action" that demands human sign-off and disclosure, or merely clerical support akin to a research attorney's memo, is unsettled. The pilots are effectively running ahead of the law that would govern them.

The criminal-justice stakes

The alarm sharpens where liberty is on the line. The Los Angeles contract permits the tool's use on post-conviction relief and, according to judges who attended an internal presentation, raised the prospect of assisting with petitions under California's Racial Justice Act, which lets incarcerated people challenge convictions tainted by racial bias.

Los Angeles County District Attorney Nathan Hochman called those criminal motions his "greatest concern." "When you're dealing with someone's liberty — as opposed to in the civil setting, which is everything other than liberty — the stakes couldn't be higher," he said. "I don't want to take the chance, particularly in a criminal case, that AI happens to get it wrong. And now someone's constitutional rights have been infringed."

One Los Angeles judge, speaking anonymously, was blunter: "I think it is outrageous. AI cannot and never will be able to replace human judgment in evaluating complex social dynamics. Ultimately, that will erode the public's confidence in the competence and fairness of the judiciary." Klapper and Slayton both said the court has no current plans to apply the tool to Racial Justice Act petitions.

Why it matters

The pilot is one node in a broader wave of government AI adoption — Ohio, for instance, has moved toward AI-assisted screening in benefits-eligibility determinations — testing how far automated systems can reach into consequential decisions before the rules catch up. The judiciary is the highest-stakes test of all, because a court's legitimacy rests on litigants believing a human weighed their case.

That makes the "litigants won't know" posture the likely flashpoint. A defendant who learns, after the fact, that an undisclosed AI shaped the ruling against them has a ready-made due-process argument, and a single such appeal could force California's courts to confront whether AI review creates a disclosure duty. The answer would not stay in California. As the first major state-court system to wire commercial AI into judicial drafting, California is, by default, writing the precedent for AI transparency in American courtrooms — even if it is doing so without telling anyone whose case is in the queue.