A Master Table of Truth: Lawyers Using AI (craigball.net)

After a string of embarrassing court filings in which large language models produced entirely fabricated case citations, a practical fix has been proposed: build a publicly accessible “master table” of bona fide reported decisions that acts like a NIST NSRL for case law. Integrated into AI-assisted drafting tools, the table would let software check every citation—case name, reporter, volume, page, court and year—against a canonical list and flag anything missing for human review. The goal is straightforward: stop hallucinated citations from ever reaching judges or dockets and reduce the malpractice and credibility risks of careless AI use. Technically, the idea raises clear challenges but remains feasible. The table would need continuous updates, wide coverage (federal and state appellate decisions, published and many unpublished opinions), normalization across multiple reporters and citation formats, and robust parsing tolerant of punctuation errors. It wouldn’t be a silver bullet—models could still misstate holdings or use genuine cases out of context—but it would dramatically cut the most egregious fabrications. Governance is key: proprietary vendors already hold much of this data, while projects like the Caselaw Access Project and CourtListener could provide open foundations; coordination by institutions (NIST, Library of Congress, law libraries) or a consortium would be required to maintain standards and trust.