Source: Thomson Reuters

Judges across the United States are adopting generative AI tools to augment judicial work—improving clarity, efficiency, and access to justice—while explicitly avoiding automation of decision‑making. A webinar hosted by the AI Policy Consortium (a joint effort of the National Center for State Courts and the Thomson Reuters Institute) featured Prof. Amy Cyphert and U.S. Magistrate Judge Maritza Dominguez Braswell, who described early adoption experiences and emphasized deliberate, informed use. Practical applications described include translating complex legal reasoning into plain language for hearings, offering plain‑language summaries to self‑represented litigants, producing explanatory videos and avatars, and helping litigants navigate court processes.

Adoption is governed by strict guardrails and peer‑led education: segregation of work and personal accounts, prompt anonymization, prohibitions on clerks uploading sensitive material, training on high‑risk vs low‑risk use, structured prompts, and standing instructions. One concrete project uploaded a district’s local rules, the Federal Rules of Civil Procedure, and standing orders and instructed the tool: don’t answer unless grounded in a rule, cite the rule with every response, and say if you don’t know. Speakers flagged deeper risks beyond hallucinations—bias, cognitive deskilling, and erosion of public trust—and gave examples such as a subtle defense‑side tilt when training material skews toward larger firms’ publications. Both presenters rejected mandatory disclosure requirements as chilling, instead favoring voluntary transparency and peer forums like the judge‑only Judicial AI Consortium; Prof. Cyphert also warned about younger lawyers relying too heavily on AI and missing critical learning steps.