AI Regulation Reference: A Structured Database of AI Compliance Obligations Across Jurisdictions

2 min read

If your organization builds or deploys AI systems, you are already subject to overlapping regulations from multiple jurisdictions — and the list is growing. Figuring out which rules apply, what they require, and when enforcement begins means cross-referencing dozens of documents that use different terminology for the same obligations.

AI Regulation Reference is an open-source, structured database that tracks compliance obligations across 12 regulations spanning the EU, US federal agencies, and six US states. Instead of organizing by regulation, it uses an obligation-first ontology: stable compliance concepts like transparency, human oversight, and bias prevention serve as anchors, with specific regulatory provisions mapped to them.

Quickstart

git clone https://github.com/snapsynapse/ai-regulation-reference.git
cd ai-regulation-reference
node scripts/build.js

This generates the static site and JSON API under docs/. To validate cross-references in the ontology:

node scripts/validate-ontology.js

To launch the MCP server for AI assistant integration:

node scripts/mcp-server.js

When to use it

You need a single structured source to check which AI regulations apply to your organization, compare obligations across jurisdictions, and track enforcement deadlines. The JSON API and MCP server make it practical to integrate compliance lookups into developer tooling or AI-assisted workflows rather than maintaining a spreadsheet.

When not to use it

This is a reference tool, not legal advice. It does not replace regulatory counsel and does not cover every regulation globally. If you need binding compliance assessments or jurisdiction-specific legal interpretation, consult an attorney. The current coverage is weighted toward US state laws and the EU AI Act.

What it covers

  • 12 regulations: EU AI Act, Colorado ADMT, California CCPA ADMT, CMS Medicare Advantage, and others across 8 regulatory authorities
  • 10 core obligations: transparency, human oversight, risk assessment, bias prevention, incident reporting, and more
  • 3 access layers: HTML site for browsing, JSON API for integration, MCP server for AI assistants

The obligation-first structure makes an interesting bet: that the compliance concepts are more stable than the regulations themselves. If you work in AI governance, curious whether that matches your experience.