The Law Should Be Readable by the People It Governs

In a democracy, the law is meant to express the will of the people. Yet how many people can meaningfully examine it?

The text may be public. Understanding it is another matter. Volume, specialized language, scattered sources, amendments, and cross-references create a substantial barrier between having access to the law and knowing what it says.

That gap has always bothered me. AI gives us a chance to narrow it dramatically.

I recently came across Global Law, a project that organizes legislation using Git—the version-control system familiar to software developers. It records legal texts and their changes over time, and includes an MCP server through which AI tools can query them.

Its coverage is incomplete and varies by jurisdiction. But the direction is exciting: make the law easier to examine, including how it became what it is today.

From available to accessible

In How Law and Programming Are Related, I explored the parallels between the two fields. Both use statements, conditions, terms, and rules to govern behavior. Their shared vocabulary reflects a deeper resemblance.

This project gives that comparison a practical form.

Version control lets us inspect changes instead of trying to reconstruct them from a pile of documents. Which words were added? Which were removed? What did a provision say at a particular time? Which legislation changed it?

Git does not certify the accuracy of the legal record. The sources and the care taken in reconstructing them still matter. But it provides a useful structure for examining that record.

AI adds another layer of accessibility. A person can ask a question in ordinary language, then use an assistant to locate provisions, follow references, compare versions, and explain what it found. The answer can lead back to the text and its sources, allowing the person to question the explanation.

The potential extends well beyond saving lawyers time. It could help the people governed by the law become better informed participants in it.

Public access becomes more meaningful when people can actually use it.

Is it a bug—or a safeguard?

Making legal texts easier for AI to examine also opens the way to searching for inconsistencies and unexpected consequences.

Some researchers describe this as finding legal “vulnerabilities” or “exploits.” I understand the analogy, but the language deserves care.

An apparent loophole might be an oversight. It might also be a deliberate exception, a protection against excessive state power, or an ambiguity that has not been settled. A surprising interpretation might simply be wrong.

Calling something a vulnerability can skip the very question we need to examine: what makes this result a defect?

Even legislative intent may offer no simple answer. Laws emerge from compromise. Different people can support the same text for different reasons. And a rule can work exactly as intended while still deserving reform.

The useful contribution is to expose the consequence and make the reasoning inspectable. Show the provisions, explain how they interact, and identify what remains uncertain. Then we can discuss whether the result should stand.

An opportunity to refactor

This is where the possibilities become much larger.

If AI can help us examine legislation across provisions and through time, it could support a sustained effort to improve the body of law itself.

In software, refactoring means changing the structure while preserving the intended behavior. Something similar is worth considering for legislation: clearer organization, consistent terminology, fewer redundant provisions, and references that people can follow.

Legal refactoring would require particular care. Wording carries interpretive history. A change that appears cosmetic can affect meaning. Preserving the existing effect must be demonstrated and examined, rather than assumed.

But that difficulty should encourage better tools and review, not resignation.

We could use AI to help identify provisions that appear to conflict, definitions that drift between statutes, obsolete references, and layers of amendments that have become difficult to understand. Proposed changes could be examined against concrete scenarios: who gains a right, who loses a protection, whose obligations change?

Some work would improve clarity while preserving policy. Other work would require a substantive choice to amend or replace it. Keeping that distinction visible would make a broad reform effort more accountable.

I can imagine a serious public push to review and amend accumulated law—one made more feasible by reducing the effort required to understand how its parts fit together.

The people still decide

As I argued in the earlier article, law needs room for human circumstances. Perfect mechanical consistency is not its highest purpose. Fairness can require exceptions, discretion, and judgment.

AI does not settle those choices. It can help bring them into view.

Nor does finding an inconsistency authorize anyone to rewrite the law. Reform must pass through the institutions and procedures that give it legitimacy. Better analysis supplies those processes with better questions and evidence.

That is what excites me about this development. More people could understand the rules that govern them, examine how those rules interact, and participate in improving them.

A law being published is a starting point. A law being practically accessible to the people it governs is a much more demanding—and worthwhile—goal.