I built an audited API for AI law (US + EU + global), with an MCP connector

AI regulation now changes faster than any team can track manually. The EU's Digital Omnibus just reshuffled the AI Act — and the gap between the headline and what's actually in force is exactly where teams over-comply or get exposed. (Concrete example: Article 50 transparency duties still take effect on 2 August 2026 despite the reshuffle.)

So I built AI Law Tracker — one API for AI law, designed to live inside other products.

What's in it

• Coverage: 50 US states + DC + federal + EU + other jurisdictions
• Refreshed daily, provenance + official source URL on every record
• An interpreted layer: obligations, penalties and effective dates — not just raw statute text
• Human-audited and sourced, not LLM-generated. A hallucinated deadline is worse than no answer, so the interpreted data is verified, not generated.

The part I care about most: trust

• Public accuracy ledger — every record is checkable.
• Bug bounty for wrong data.

If you're building anything downstream of the regulation layer (AI governance, compliance tooling, legal engineering), you shouldn't have to trust a black box.

MCP connector

There's a Model Context Protocol connector with 24 tools, so an agent inside Claude or ChatGPT can query laws, obligations, penalties and deadlines directly.

Try it (no card)

• Free API key: ai-law-tracker.com/developers#get-key
• Live Omnibus breakdown: ai-law-tracker.com/omnibus

I'd love feedback on the data model and the MCP tool design — is 24 tools the right granularity for an agent reasoning over regulation, or too fine-grained?