This is a likely valid application of AI. Consider that legislative text is often obfuscatory rather than transparent, wildly over-specific in some areas and with yawning loopholes in others. Contractual and judicial texts have the same problem. In some countries, legislative drafting is a task reserved to the very best lawyers (not necessarily the best people, but that's another story). In others it falls to the least capable, and they are relegated to interpreting the content sufficiently for a politician to advocate on it, although the text has often been written by a lobbyist or counsel in a case.

A first application for AI would be to take existing legal texts and redraft them for simplicity and clarity, which could be approached incrementally. This would of course expose a lot of inconsistencies and garbage, and cause much harrumphing about the (supposed) intent of the original authors. A second would to evaluate legal texts for general semantic consistency across time and code structure. A third would be to prune and consolidate legal codes in order to reduce their complexity and the burden of compliance.