Los Angeles is piloting an AI program to help judges cut through a flood of paperwork — and the experiment could matter to any sector that’s starting to see legal volume spike, including crypto.
Why it’s happening
Courts worldwide are overwhelmed by rising caseloads, and Los Angeles is no exception. A February 2026 report from law firm Fisher Phillips found filings jumped 49% in one year — from about 4,100 to 6,400 — a surge industry observers link in part to how AI lowers the cost and ease of producing filings. The Los Angeles Superior Court has launched a pilot that gives a small group of judicial officers access to an AI system called Learned Hand to test whether machine assistance can speed routine work without compromising judges’ authority.
What Learned Hand does
Learned Hand, founded in 2024 and named after the influential federal judge, is built to handle the administrative side of civil cases: summarizing filings, organizing evidence, and generating draft rulings. The aim is not to replace judicial decision-making but to reduce “drudge work” so judges can focus on legal analysis and discretion. According to founder and CEO Shlomo Klapper — a former law clerk for the U.S. Court of Appeals and a deployment strategist at Palantir — the tool covers a case end-to-end, from intake to draft rulings, and is designed to be easy to use: “It’s point and click. They don’t have to do any prompts.”
Balancing help with safeguards
The court’s presiding judge, Sergio C. Tapia II, framed the pilot as a cautious experiment: while the tool could change how officers review files, it will not “replace, or in any way compromise, the sanctity, independence, and impartiality of judicial decision-making.” Klapper emphasizes judges must verify any AI output and not take it at face value: “Don’t trust, verify. They shouldn’t trust anything. It has to show its worth.”
Tackling the hard problem: verification, not generation
Klapper argues the technical challenge is not getting models to produce text — that’s comparatively easy — but ensuring those outputs are reliably tied to the underlying case materials and legal authorities. “Most of the expense of our large language model is in the verification, not the generation,” he told Decrypt. To reduce the risk of false or misleading outputs (so-called hallucinations), Learned Hand restricts its knowledge base to a defined set of legal materials rather than scraping the open web, and it breaks tasks into discrete steps, assigning each to a model with a specific function.
Real-world cautionary tales
Concerns about AI hallucinations aren’t hypothetical. In 2023, defense attorneys in the Pras Michel case alleged an AI had helped craft a closing argument that included frivolous claims and missed weaknesses. That same year, a federal judge ordered lawyers for Michael Cohen to provide printed copies of cited cases after the court couldn’t verify them online. Those incidents have amplified calls for stricter guardrails when AI touches legal work.
Why this matters to crypto readers
For projects, firms, and builders in the crypto space, the pilot is a useful bellwether. As AI lowers legal costs and enables higher filing volumes, courts will need scalable tools to keep up with disputes ranging from securities allegations to smart-contract litigation. Systems like Learned Hand, if proven reliable, could speed case processing and change how legal teams prepare and present material — but they also force new questions about verification, sources, and institutional trust in automated aids.
Bottom line
The Los Angeles pilot is a measured attempt to use AI to reduce administrative burdens on judges while preserving human control over legal decisions. The experiment highlights a core truth about legal AI: generation is the easy part; rigorous verification and carefully controlled source materials are where the real work — and cost — lie. As courts and industries adapt to both AI-generated filings and the opportunities these tools offer, “don’t trust, verify” may become the new mantra for legal technology adoption.
Read more AI-generated news on: undefined/news
