Regulatory Intelligence from a Living Legal Corpus

By Anushka Appala and Dr. Janio Rosales

To make a government decision defensible, NaciluzIA had to solve a problem that looked mundane and turned out to be foundational: the law had to become something a machine could read, cite, and version. Statutes, regulations, and administrative rules are written for lawyers and published as prose; an automated system that grounds its answers in "the law" needs that law as structured, addressable, version-controlled text. Building that corpus was a precondition for the platform. It also turned out to be a product in its own right — and the audience for it is not only the State.

Any company operating in Guatemala, and increasingly across Central America, lives inside the same body of rules a ministry does: tax codes, labor law, customs and trade rules, sector-specific regulation. Compliance teams navigate this the hard way — reading PDFs, tracking amendments by hand, and hoping the version they relied on last quarter is still current. It is slow, it is error-prone, and when it goes wrong the cost is not academic. The same versioned legal corpus that lets NaciluzIA cite the exact article behind a citizen's decision can answer a company's compliance question with the same rigor.

Grounded, cited, and versioned — the three properties that matter

A generic chatbot pointed at the law is worse than useless in a compliance setting, because a confident wrong answer about a filing obligation is a fine or a shutdown, not a bad search result. What makes the corpus a compliance instrument rather than a novelty is three properties.

It is grounded: every answer is retrieved from the actual legal text, not generated from a model's memory, so it can be traced to source. It is cited: the answer names the exact article, decree, and clause it rests on, so a human expert can verify it in seconds instead of re-deriving it. And it is versioned: because the corpus tracks the law's history, the system can answer not just "what is the rule?" but "what was the rule on the date this transaction happened, and what changed since?" That last property is the one compliance teams never have and most need.

        COMPLIANCE QUESTION
   "What are the withholding obligations
    for this contract, as of March 2025?"
                │
                ▼
      ┌───────────────────────┐
      │  RETRIEVAL over the    │
      │  VERSIONED LEGAL CORPUS│
      │  (article-level,       │
      │   date-aware)          │
      └───────────────────────┘
                │
                ▼
   ANSWER  ──►  plain-language response
        │
        ├─►  CITATION: Decreto ##-####, Art. ##  (v. in force 2025-03)
        ├─►  CHANGE LOG: amended 2025-07 — what differs
        └─►  CONFIDENCE: grounded in text, flagged if ambiguous
                │
                ▼
        HUMAN REVIEW where it matters

Nothing here asks a company to trust a model. It asks them to trust the law — and gives them a faster, checkable path to it.

A public good and a product, without contradiction

This is where the open-standard model and the commercial model meet without conflict. The versioned corpus of national law is a public good: it should be open, machine-readable, and citable by anyone — a piece of civic infrastructure that makes the whole legal system more legible. That is how it is stewarded, and keeping it open is a promise, not a limitation. What institutions pay for is not the law — it is the integration: connecting the corpus to their own systems, tuning it to their sector, keeping their obligations current, and supporting the teams that rely on it. The standard stays open; the service is where the sustainability comes from.

The market for exactly this is large and growing. RegTech is heading toward roughly $77B by 2034, and "regulatory intelligence" — keeping an organization current with the rules that bind it — is one of its named segments (IMARC, 2025). The AI-governance market growing alongside it, at about 36% a year (Grand View, 2025), is driven by the same force: regulation is multiplying, and the cost of getting it wrong is rising. A tool that turns a jurisdiction's law into a queryable, cited, version-aware resource is useful to a bank's compliance office, a manufacturer's customs team, a professional-services firm advising clients, and a foreign investor sizing up country risk.

The same asset, two audiences

There is an efficiency here that is easy to miss. The corpus does not need to be built twice. The work required to make Guatemalan law machine-readable so a ministry can ground a decision is the same work that makes it useful to a company that must comply with that law. Government funds and legitimizes the public good; the private sector sustains it commercially; and as NaciluzIA replicates — El Salvador next, then the wider region — each new jurisdiction's corpus serves both audiences from a single build. The law becomes infrastructure, and infrastructure, done in the open, is something both a state and a market can stand on.

Cada decisión, a la luz.

— Anushka Appala and Dr. Janio Rosales