Git-Style Version Control for US Federal Law
Original: Git for the US Code
Why This Matters
LLMs entering legal research makes stale-statute errors a live liability risk, not an academic one.
Lexifina argues US federal statutes need a version-controlled system like Git, noting that amendment history is buried in prose notes and language models frequently apply superseded rules when answering legal questions.
Legal tech company Lexifina published an opinion piece arguing that US federal law should be tracked with the same version-control precision that software engineers apply to code — showing who changed a line, when, and what the text looked like on any given date. Currently, the Office of the Law Revision Counsel publishes the US Code as a series of release points tied to named public laws, with amendment history recorded in prose notes and effective dates scattered across separate sections.
The piece uses 11 USC § 1182(1) — the small business debtor definition — as a case study in how tangled this gets. The debt limit changed in 2019, spiked to $7.5 million under the CARES Act in 2020, lapsed in March 2022, was retroactively restored in June 2022 for pending cases, expired again in June 2024, and now sits at $3,424,000 after an April 2025 adjustment. A lawyer relying on the release point printed at filing time could get the wrong number.
The argument sharpens when AI enters the picture. A Technical University of Munich study found that language models answering questions on German statutes applied superseded rules and preferred newer provisions even when older law governed. One model scored 24% on post-cutoff questions unaided, jumping to 87% when the correct statutory text was retrieved. The author concludes that temporal validity must be treated as a hard constraint, not a prompt-engineering suggestion.