drdavidprivacy/us-ai-law-database-census

US AI Law Database, census

Release v0.8.0-2026-11-07. Published by Dr. David, LLC for the AI Governance Collective.

What this is

This repository holds the public census of the US AI Law Database: one row for every AI-specific United States statute the library tracks, with its identity (state, name, bills, citation), its dates (signed, effective, operative, sunset), its status, the regulatory-model band it belongs to, and a link to the official source. The census is the index of the database, published so that anyone can see which laws it covers and check the counts.

The full database sits behind membership at https://laws.aigovcollective.com. Each full record carries the verbatim statutory text, covered roles and the statute's own actor terms, thresholds, enforcer, private right of action, penalties with units, cure periods, rulemaking delegations, litigation posture, and political metadata. Across the release the full database holds 524 typed duties, 378 verbatim definitions, 26 rulemaking delegations, and 235 roll-call rows, and every field carries a confidence label. None of that is in this repository; only the census fields listed below are.

Two files carry the same 125 rows: census.csv and census.json. In the JSON file band is an integer, counted is a boolean, and empty dates are null.

How it was built

The database rests on a three-level source hierarchy. First, the enacted statutory text, transcribed verbatim from the official legislature or code site or, where that site blocks automated retrieval, from a mirror with a dual-source cross-check on every operative passage. Statutory words are never edited; drafting quirks are preserved and noted. Second, a private library of 125 Markdown law notes, one per instrument, each holding the full verbatim text and a section-cited summary. Third, one YAML record per law, authored from the notes, with every legal-fact field citing its note section and block anchor. Political fields cite legislature and vote records directly.

Every field in the full database carries one of five confidence labels: verified-verbatim, dual-sourced, single-sourced, editorial, or pending. The band assignment is editorial; the dates, status, bills, and citation in this census come from the statutory text and legislative record.

Every batch of notes and every release goes through an adversarial fact-check: an independent reviewer is instructed to refute each claim against the verbatim text, and findings are graded fatal, material, or minor and corrected before release. The audit for v0.8.0 used two reviewers and returned 44 findings, none fatal, all corrected. The v0.7.0 audit (October 11, 2026) used five reviewers and returned 95 findings (4 fatal, 39 material, 52 minor), all corrected. The v0.6.0 audit (September 13, 2026) returned 47 findings (0 fatal, 8 material, 39 minor), all corrected.

A single compile script is the only path from the YAML records to the published outputs; derived files are never hand-edited. The compile validates every record against a schema, checks that every cited note and anchor exists, and runs a reconciliation gate that fails the build if the count of current instruments disagrees with the library's index. Each release is a frozen snapshot with a numbered, dated tag, and every row in this census carries that tag in its release field.

The eleven bands

Bands group laws by regulatory model, by what the law regulates and whom it binds. Band assignment is an editorial judgment and is labeled as such in the full database.

1. Frontier Safety: laws that bind developers of the largest AI models to safety frameworks, incident reporting, or similar duties tied to catastrophic risk.
2. Consequential Decisions: laws governing AI or automated systems that make or support decisions with legal or similarly significant effects on people, including employment and workplace surveillance.
3. Transparency: laws requiring disclosure that content is AI-generated or that AI is in use, or provenance and training-data disclosures, outside the election and intimate-imagery contexts.
4. Interaction and Companions: laws governing chatbots and AI companions that converse with users, including duties toward minors and crisis protocols.
5. Health Care: laws governing AI in clinical care, utilization review and prior authorization, and patient-facing communications.
6. Synthetic Media and NCII: laws on election deepfakes, non-consensual intimate imagery, digital replicas, and synthetic performers; the one federal instrument sits here.
7. Omnibus: a single act that imposes duties across several of the other bands at once.
8. Algorithmic Pricing: laws restricting prices or rents set by algorithm, including rent-setting software and pricing based on personal data.
9. Personhood and Legal Status: laws declaring that an AI system is not a person, cannot hold a role reserved to natural persons, or cannot stand in for a human in a defined position.
10. Audit and Assurance: laws governing AI auditors and independent verification bodies, where the duty runs to the assurer.
11. Professions and Public Institutions: laws governing how licensed professionals, courts, and public institutions may use, disclose, or procure AI, where the duty runs to the professional or institution rather than to a developer or deployer.

Fields

Field	Description
instrument_id	Stable identifier for the instrument, lowercase, state prefix plus bill or code reference (ca-sb53, ny-gbl-art48). Shared with the full database.
state	Two-letter USPS code of the enacting state, or US for the one federal law.
name	Short editorial name: state code, bill or code reference, and subject.
band	Band number, 1 through 11.
band_name	Band name matching the number (see the list above).
status	Legal status at release; see the status vocabulary below.
counted	True when the instrument counts toward the current in-scope total; False for reference records that are excluded from the count but retained. Derived mechanically from status.
bills	The bill or bills enacted, with session, sponsor where recorded, and any substitution notes. Multi-bill instruments list every constituent bill.
citation	Session-law and codified citation as recorded; where a citation element is single-sourced or not yet posted, the field says so.
signed_date	Date the governor (or, for the federal law, the President) signed, ISO 8601.
effective_date	Date the act takes effect, ISO 8601. Where a non-urgency California act is silent, this is the constitutional default of January 1 following enactment and is labeled inferred in the full database.
operative_date	Date the operative provisions begin to apply when later than the effective date; empty otherwise.
sunset_date	Date the act or its operative provisions expire; empty when none.
official_source_url	The official legislature or code page for the statute. Mirrors are never cited here.
release	Release tag the row was compiled under (v0.8.0-2026-11-07).

A field-by-field summary with vocabularies is also in FIELDS.md.

Status vocabulary

Counted statuses (the instrument is current and in scope):

• in-force: effective and operative at release.
• enacted-not-yet-effective: signed, effective date in the future.
• operative-future: effective, but the operative provisions begin on a later date.

Reference statuses (retained, not counted):

• repealed: repealed before or after taking effect (Colorado SB 24-205).
• superseded: replaced by a later enactment, which has its own record (the original 2025 New York RAISE Act).
• struck: held unconstitutional or otherwise invalidated by a court (California AB 2839, Hawaii Act 191).
• enjoined: enforcement blocked by court order, in whole or in part, at release (California AB 2655, Minnesota Statutes section 325E.91).
• study-act: directs a study or report rather than imposing operative duties (Tennessee SB 1700).
• context-note: kept for context on a cluster of laws without itself being an AI statute of the kind the database admits (Tennessee SB 2041).

Coverage and counts, release v0.8.0-2026-11-07

• 125 rows: 117 counted current instruments and 8 reference records.
• 124 state instruments from 39 states, and 1 federal law (the TAKE IT DOWN Act, Public Law 119-12).
• Signed dates run from June 14, 2019 to October 28, 2026.
• Counted status split: 76 in force, 37 enacted and not yet effective, 4 operative on a future date.
• Counted instruments by band: Synthetic Media and NCII 45; Health Care 19; Interaction and Companions 17; Transparency 10; Consequential Decisions 8; Algorithmic Pricing 5; Personhood and Legal Status 4; Professions and Public Institutions 3; Frontier Safety 3; Audit and Assurance 2; Omnibus 1.
• Newest instruments: New York Chapter 368 of 2026 (General Business Law Article 48, unsafe AI companion features for minors; signed October 28, 2026, effective January 1, 2027) and Pennsylvania Act 99 of 2026 (SB 806, AI deceptive-advertising disclosure; signed October 22, 2026, effective December 21, 2026), the first Pennsylvania entry.

Scope: enacted statutes only. No regulations, executive orders, or pending bills. All fifty states were screened for enactments from 2019 through October 2026. Election-deepfake laws are in; general privacy laws and neurorights laws are out. A federal statute enters only where it anchors a state cluster. Within the scope it claims the set is treated as complete at the release date, but it is a curated set, not a census of every statute that mentions AI, and a law absent from this file is not evidence that no law applies.

Limitations

• This is a reference for governance work, not legal advice. Verify against the official source linked in each row before relying on it for a legal position.
• Dates can lag the record. Effective and operative dates are transcribed at release and are not re-checked between releases; a court order, amendment, or delayed posting after November 7, 2026 is not reflected until the next release.
• Every 2026 California text was mirrored from LegiScan because the official site blocks automated retrieval, and California chapter numbers for 2026 acts come from aggregators pending official posting. All seventeen sit in a verification queue for a check against the official host.
• Two court orders are described from press reports because the orders themselves were not retrieved: RealPage v. James (S.D.N.Y., September 29, 2026), which enjoined part of New York General Business Law section 340-b, and X.AI v. Ellison (8th Cir., about October 2, 2026), which enjoined Minnesota Statutes section 325E.91 pending appeal.
• At release, the New York Assembly's roll-call tally for Chapter 368 of 2026 had not been published, and Pennsylvania had not posted the enrolled text of Act 99 of 2026; the Pennsylvania record was extracted from the final printer's number of SB 806 and its Pamphlet Laws page citation is single-sourced.
• Band assignment is an editorial classification, not a statutory fact.
• Effective is not operative. The two dates are separate columns for a reason; read both.

Update cadence

Legislative scans run monthly. Each scan that finds new enactments or needed corrections produces a numbered, dated release; seven releases shipped between August and November 2026. The next scheduled pass is January 2027, when the amended texts of California SB 942 and AB 853 (as amended by SB 1000 and AB 2713) swap in. Every release is a frozen snapshot, and this repository is updated to match each release. Corrections are logged per release in the database's provenance record.

License

This census (census.csv, census.json, FIELDS.md) and the project datasheet are licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). The statutory text the database quotes is public domain. The full database at https://laws.aigovcollective.com is available under member terms (view, search, cite with attribution, and use in your own organization's compliance work; no bulk download, scraping, redistribution, login sharing, resale, or model training). Compilation copyright 2026 Dr. David, LLC.

Citation

Dr. David, LLC. US AI Law Database, census release v0.8.0-2026-11-07. https://laws.aigovcollective.com

A CITATION.cff file is included in this repository.

Contact

Questions and corrections: laws@aigovcollective.com. Every correction that survives fact-check is logged and shipped in the next numbered release.

Maintained by Dr. Kyle David, founder of the AI Governance Collective.