A live, quantitative map of the global AI-regulation landscape — built entirely from our own primary-sourced records, recomputed from the database on every load.
AI regulation is no longer a single story — it is being written across 90 jurisdictions with binding AI law at once, alongside 16 jurisdictions with a voluntary AI framework (a separate shelf that overlaps this one — a model governance framework is real coverage but is not law, so the two are shown side by side and never added). This benchmark counts every binding AI-specific law and bill we track — 1,402 records in all — and breaks them down by where they sit, when they landed, what they regulate, and what they cost to get wrong. Every number is a live count over the same audited records that power our law tracker and public API; each record is hand-checked against a primary government source, and 100% carry that source link. Nothing here is estimated. This is a compliance-tracking resource, not legal advice.
Source: live count over AI Law Tracker's legal_records corpus (1,402 records, 90 jurisdictions with binding AI law; 16 jurisdictions have a voluntary AI framework — a separate, overlapping shelf that is listed but never summed into the binding count), editorial data last verified Jul 13, 2026.
The corpus spans four scopes. State bills dominate by volume; the EU and other national regimes are fewer laws but far broader in reach.
Source: legal_records grouped by scope and in_force. Explore each via by-status, federal, EU, and global.
Each record grouped by its primary-source date — the year the bill was introduced or the law was enacted. The concentration in the most recent years is the story: AI legislation went from a trickle to a flood.
Source: legal_records.record_date (introduction / enactment date), grouped by year. Records dated on or before 2021 are folded into one bucket; the 64 records with no single dated primary source (e.g. multi-part frameworks) are omitted from this chart only. Note: this reflects the dates on the laws themselves, not when we added them — we do not infer legislative “velocity” from our own ingestion timestamps.
Recurring themes, counted by scanning the real title and summary text of every record for each topic. A record can match several themes; the number is how many records mention it. This is descriptive keyword tagging, not a precise legal classification.
Telling people when AI is being used on them — the single most common thread.
Guardrails on AI in medical, mental-health, and behavioral-health settings.
Labeling AI-generated media, especially in political and election communications.
Disclosure and safety duties for public-facing chatbots and companion bots.
Bias audits and notice requirements for AI used in hiring, promotion, and discipline.
Duties tied to “high-risk” or consequential automated decisions, incl. anti-discrimination.
Criminalizing AI-generated CSAM and non-consensual intimate imagery.
Source: case-insensitive keyword match over legal_records.title + summary. Counts are reproducible against the public API using the q= search filter.
What it costs to get AI compliance wrong ranges enormously — from a few hundred dollars per violation to a share of global revenue. Each figure below is quoted verbatim from the law and links to its official source.
| Jurisdiction | Law | Maximum penalty (verbatim) |
|---|---|---|
| California | California AI Transparency Act (SB 942) ↗ | $5,000 per violation; each day is a discrete violation |
| Colorado | Colorado AI Act (SB 24-205) ↗ | AG-enforced (Colorado Consumer Protection Act); up to ~$20,000 per violation |
| European Union | EU AI Act (Regulation (EU) 2024/1689) ↗ | Up to €35M or 7% of global annual turnover |
| Illinois | Illinois AI in Employment Law (amends the Human Rights Act) (HB 3773) ↗ | IDHR/IHRC make-whole relief + tiered civil penalties up to ~$16,000–$70,000 per act, per aggrieved party |
| Minnesota | Minnesota Consumer Data Privacy Act (automated decisions / profiling) (Minn. Stat. ch. 325M) ↗ | Up to $7,500 per violation |
| New York | New York RAISE Act (frontier-AI safety) (S6953B) ↗ | New York DFS oversight of large frontier-AI developers |
| New York City | NYC Local Law 144 (automated employment decision tools) (Local Law 144) ↗ | $500–$1,500 per violation |
| Tennessee | Tennessee ELVIS Act (AI voice & likeness) (HB 2091) ↗ | Civil damages |
| Texas | Texas Responsible AI Governance Act (TRAIGA) (HB 149) ↗ | AG-enforced (no private right of action); up to $100,000 per uncurable violation + $40,000/day |
| Utah | Utah AI Policy Act (SB 149) ↗ | Up to $2,500 per violation (Utah Division of Consumer Protection) |
| Utah | Utah AI Mental-Health Chatbot Rules (HB 452) ↗ | Up to $2,500 per violation (Utah Division of Consumer Protection) |
Source: landmark_bills.penalty — curated, primary-sourced AI laws. Penalty text is quoted as written; see each law's full profile for scope and enforcement detail. Penalties are set by the enacting government, not by AI Law Tracker.
With no single federal AI statute, the 50 states and DC are each writing their own rules. Here is where the 51 US jurisdictions stand — the full narrative is in The State of U.S. AI Regulation 2026.
An AI-specific or AI-applicable law is enforceable today.
Signed into law with a compliance deadline on the calendar.
Bills moving through a legislature or task forces preparing recommendations.
No comprehensive state AI statute in force; federal rules still apply.
Source: jurisdiction_laws (state level), grouped by verbatim status — the same taxonomy as the by-status hub. Buckets are mutually exclusive and sum to 51.
Every figure on this page is computed live from AI Law Tracker's records at the moment you load it — there are no stored or hand-typed totals. Each record is audited against a primary government source (statute, bill text, or official regulator) and carries a verification date; records without a primary source are excluded from indexing by our source guard. Theme tagging is disclosed keyword matching over the real law text; penalty figures are quoted verbatim from the enacting law. We publish what a primary source can back and nothing more — this is a compliance-tracking resource, not legal advice.
Data last verified Jul 13, 2026. Suggested citation:
AI Law Tracker, “The AI Regulation Benchmark,” updated Jul 13, 2026. https://ai-law-tracker.com/benchmark