Session Expired

Your session has expired. Please sign in again to continue where you left off.

Sign In Again
Home › Methodology

How We Source, Verify & Update AI Legislation Data

A transparent, primary-source methodology for tracking AI laws across all 50 states. Updated continuously, reviewed by editors, and published only when confidence thresholds are met.

Bottom line Every bill we publish is sourced from official state legislature records. Bills below a 90% internal confidence score are not published. Readers can submit corrections, which are reviewed by an editor before any change ships.

1. Sources of record

We source AI-related legislation directly from the following primary records:

  • State legislature websites — the canonical record for each bill's text, status, sponsors, and committee history. Every bill page on this site links to its official_url on the state legislature's own website.
  • LegiScan — a nationwide bill-tracking API that aggregates state legislature data. We use LegiScan for collection only; final fact-checks always reference the underlying state legislature record.
  • Congress.gov — the source of record for federal AI legislation.
  • State Attorney General and agency rulemaking notices — for enforcement, regulatory guidance, and effective-date adjustments.
  • Federal Register and court records — for federal rules, executive orders, and litigation affecting state laws (e.g., preliminary injunctions on enforcement).

We do not source from secondary aggregators, social media, or AI-generated summaries of legislation.

2. Verification pipeline

Every bill passes through five stages before it becomes publicly visible:

1
CollectionBills matching AI-related keywords are pulled from official records on a daily schedule.
2
Relevance scoringEach bill is scored for AI relevance. Bills below 0.5 relevance are excluded from the directory.
3
Fact-checkSponsors, status, dates, penalty amounts, and effective dates are extracted and cross-checked against the bill's primary text. Conflicts are flagged for editor review.
4
Confidence thresholdBills with an internal confidence score below 90 are not auto-published. They sit in editorial review until a human reviews and approves (or the score improves on a subsequent run).
5
Continuous re-verificationA daily data-quality check audits the full database for duplicates, freshness, and confidence regressions. A separate twice-weekly content-freshness check re-verifies the most-trafficked pages against the underlying primary sources.
What "90% confidence" means. Our internal confidence score is a weighted composite of relevance, fact-check pass rate, and source-completeness checks. It is not a guarantee of accuracy — it's a publication threshold. Anything below it does not ship.

3. Editorial standards for analysis & commentary

Bill summaries, blog posts, and tracker pages on this site contain editorial analysis in addition to verbatim primary-source data. Our standards for that analysis:

  • Primary-source citations. Every factual claim in editorial content must link to an authoritative primary source — typically a state legislature URL, agency notice, court order, or official press release.
  • No hallucinated dates, dollar amounts, or sponsor names. Every numeric claim and named party in editorial content must be traceable to a source link on the same page.
  • Effective-date discipline. When a bill is amended (e.g., SB 942's effective date moved from January 1 to August 2, 2026 by AB 853), the editorial record is updated to reflect the new effective date, and the change is logged.
  • Status discipline. Litigation status (e.g., a preliminary injunction) is noted explicitly. We do not describe enjoined laws as "in effect."
  • Author byline. Editorial content is published under the "AI Laws by State" byline. Individual contributor names are not used.

4. What we do not claim

To be transparent about our limits:

  • We are not a law firm and we do not provide legal advice. Editorial content is informational. Compliance decisions require a licensed attorney.
  • We do not claim to capture every AI-related bill in real time. There is typically a lag between a bill's introduction and its appearance in our directory while it is verified.
  • We do not claim our analysis is exhaustive. Editorial content covers what we believe is most useful to compliance professionals, not every clause of every bill.
  • We do not claim our quality threshold is perfect. A bill that passes our 90% threshold may still contain errors. That is what the corrections process is for.

5. Corrections process

If you find an error on this site — a wrong date, a misstated penalty, a stale status, an incorrect sponsor — please tell us. Every submission is reviewed by an editor. If we agree, we publish a correction and update the underlying record. We log accepted corrections publicly so readers can see what changed and when.

Found something wrong?

It only takes a minute. We respond to every submission with a source.

Submit a correction

6. Update cadence

  • Bill collection: daily, from official records.
  • Data quality audit: daily, with notification on any duplicate, freshness, or confidence regression.
  • Content freshness review: twice weekly, focused on the highest-traffic pages.
  • Editorial publication: as new analysis is needed; every blog post is timestamped and re-verified on the freshness cycle.

7. Disclosure: technology stack

This site uses automated tools at multiple stages of the pipeline — including for collection, relevance scoring, and draft fact-extraction. Automation does not replace editorial review. Final publication decisions on editorial content are made by an editor, and the 90% confidence threshold gate on bill publication is a hard cutoff — nothing below it ships automatically.

We disclose this because we believe readers should know how the data they read is produced.

Reach an editor

For corrections, please use the corrections form. For press inquiries, partnerships, or questions about our methodology, email [email protected].

Last updated April 30, 2026. This page is maintained by AI Laws by State and is updated whenever our process changes.