Data sources & methodology

Where RegIntel's regulatory data comes from, how it's collected, how to verify it against the original source, and what we don't cover. Written for developers, compliance teams, and AI applications that need to know they can trust the data.

What RegIntel actually contains

RegIntel is a structured database of 212 regulations across 41 jurisdictions, each represented as a JSON record with the following fields:

• id — stable identifier (e.g. 123 (integer))
• name and jurisdiction — what the rule is and where it applies
• category and tags — taxonomy (privacy, AML, payments, crypto, etc.)
• obligations — what the regulated party is required to do
• penalties — quantified consequences for non-compliance
• scope — who the rule applies to
• key_articles — pointers into the underlying legislation
• source_url — link to the authoritative regulator page
• last_updated — when the record itself was last reviewed

The source_url field is the load-bearing one for trust. Every record points to the regulator's own page — the data is a structured rendering of what they publish, not original interpretation.

Source regulators

The data is collected from official regulator publications. Major sources include:

The complete list with regulation counts per jurisdiction is available via GET /jurisdictions — that endpoint does not consume credits.

How records are collected and structured

The pipeline is in three stages:

1. Acquisition. Updater scripts (open source at ad0750/regintel-mcp for the public surface; private for the ingest pipeline) pull from regulator websites, official PDFs, and government RSS/news feeds. Each source has its own parser because regulator websites vary widely in structure.
2. Structuring. Raw text is normalised into the JSON shape above. Obligations and penalties are extracted into discrete fields rather than left as free text.
3. Linking. Every record retains its source_url so downstream consumers can verify against the regulator. We treat broken source URLs as a bug — if you find one, mail support@regintelapi.com and we'll fix it.

How to verify any record against the source

Two-step verification, designed to be runnable by both humans and LLMs:

# 1. Get the record
curl "https://api.regintelapi.com/regulations/123" \
  -H "x-api-key: YOUR_API_KEY"

# 2. Follow the source_url in the response, read it, compare

This pattern is especially important for AI/LLM applications. When an LLM uses RegIntel to answer a regulatory question, the answer should include the source_url as a citation. The user can then click through to the regulator's own page — the LLM's claim is verifiable, not a hallucination. The For AI page walks through the citation pattern in more detail.

Freshness and update cadence

Updater scripts run continuously to pick up regulator announcements. Most amendments and new regulations appear in the API within 24–48 hours of public release at the source. The GET /updates?since=<date> endpoint exposes the modification timeline so downstream caches and vector stores can sync incrementally.

We don't claim real-time. Regulators themselves often publish on rolling timelines (a regulation can be "in force" before the regulator's website fully reflects it), so any product claiming "real-time regulatory data" is overpromising. Our commitment is: structured, source-cited, and corrigible within 48 hours of any discrepancy being reported.

What RegIntel does NOT cover

Being explicit about scope is part of being trustworthy. RegIntel does not currently include:

• Sub-national regulations in most jurisdictions (US state-level AML rules, German Land-level data protection, etc.) — except where they are widely cited or unusually material.
• Sanctions lists, PEP databases, adverse media data — that is a different product category. For screening use a dedicated provider; see our RegIntel vs ComplyAdvantage page for the distinction.
• Industry self-regulatory codes (e.g. trade-body codes of conduct) unless they have force of law in a jurisdiction.
• Editorial commentary or impact analysis — RegIntel gives you the structured data, not interpretation. For analyst-written commentary, see enterprise alternatives.
• Predictive or forward-looking signals — we report what exists, not what we think regulators will do next.

Errata and corrections

If you find a record that disagrees with its source_url, or a regulator whose latest amendment hasn't propagated, email support@regintelapi.com with the regulation ID. We treat regulator-disagreement as P1 and aim to correct within 48 hours.

This is not legal advice