Developer builds Claude Code tool to audit global legislation
Developer SnooPeripherals5313 says they used Claude Code to build a free-to-use tool that tracks legislation through Git blame, version history, timelines, and expanding sets of detectors. The developer says it mirrors 107 legal corpora read-only and indexes about 4.3 million historical file changes in DuckDB, and that its engines surfaced alleged loopholes and logical errors in laws.
The project presents a concrete AI-assisted approach to making legislative changes traceable and flagging potential drafting errors, although the developer acknowledges that many findings may be unenforceable in practice.
AI for cybersecurity is taking the world by storm, leading to the need to harden systems all over software.
I built this project with Claude Code. It's completely free to use/try. The approach is pretty simple: use claude to make legislation attributable (git blame, versioning with timelines). Claude Code (Claude Opus 5.5) wrote nearly all the code. I set the goals, chose the sources and made the design decisions. Specifically: it mirrors the 107 corpora of the global-law repo read-only, indexes about 4.3 million file changes from their git histories into DuckDB, and adds a growing set of detectors.
But we have access to the same engines for law. Our global law repo keeps a versioned history of law/caselaw across the world. So, we pointed AI at it, and found all sorts of loopholes and logical errors.
Pah, you say. Most of this is unenforcable in reality. And you might be right. Nonetheless, we all know that legal loopholes are a part of reality. And it would be nice to rewrite errors, even when the errors are unenforcable.
As part of our monitoring system, where we're constantly checking official legislation for upcoming laws, we now also check for vulnerabilities and mistakes.
In an ideal world, the jurisdiction in question could open a pull request and sort this out.