U.S. AI Laws by State 2026
By Axis Intelligence Research
Co-author: Sarah Mitchell | Last updated: July 30, 2026 | License: CC BY 4.0
Quick Answer
State AI regulation in the US has moved from fringe to mandatory compliance terrain. As of July 2026, at least 5 states have comprehensive AI laws either in force or effective January 1, 2027 — California, Colorado, Texas, New York, and Illinois — alongside a patchwork of 40+ narrower laws on deepfakes, hiring, and chatbots. 145 AI bills became law across all 50 states in 2025, the first year every state introduced at least one. The federal moratorium that would have frozen this activity was stripped from the One Big Beautiful Bill Act by a 99–1 Senate vote in July 2025, leaving states in full control of the regulatory agenda.
Key Findings
- State AI bills more than sextupled in two years: lawmakers introduced fewer than 200 AI-related bills across all state legislatures in 2023; that number reached 635 in 2024 and 1,208 in 2025, per MultiState — a 504% cumulative increase in two legislative cycles, according to Axis Intelligence Research’s cross-source calculation (using 200 as the 2023 upper-bound figure).
- 145 state AI bills became law in 2025, up from 99 in 2024, according to MultiState — the largest single-year enactment total recorded for AI-specific legislation in the United States.
- California leads state AI regulation stringency: Axis Intelligence Research’s ASCSI™ score (see Section 8) ranks California at 9/10 and New York City at 9/10, with Colorado, Texas, and New York each at 7/10 — three states with different regulatory philosophies arriving at the same composite score.
- The 10-year federal moratorium is gone: the US Senate voted 99–1 on July 1, 2025 to strip the AI preemption provision from the One Big Beautiful Bill Act, according to Global Policy Watch — the closest thing to a bipartisan consensus on AI policy the 119th Congress produced.
- The 2026 bill pipeline already exceeds 2024’s full-year total: MultiState tracked 1,561 AI bills across 45 states by March 2026 alone — a 146% increase over all 635 bills introduced in all of 2024 — with sessions still active.
How Many US States Have AI Laws?
Counting state AI laws is harder than it sounds, because the answer depends entirely on what “AI law” means. Three reasonable definitions produce three very different numbers.
The broadest definition — any enacted measure that touches AI, from budget appropriations to deepfake criminal penalties — yields over 300 laws across 48 states through 2025, per the NCSL Artificial Intelligence Legislation Database, which tracks both enacted and introduced measures beginning in 2024.
A narrower definition — laws that impose affirmative obligations or penalties on organizations deploying AI in consequential settings — produces a much smaller list: as of July 2026, roughly 7 to 10 jurisdictions depending on how the threshold is drawn.
The most useful frame for compliance teams: comprehensive laws targeting private-sector AI in high-stakes decisions. By that standard, the count is 5 states with enacted frameworks, plus New York City with its own municipal law. That is the definition this article uses.
The IAPP US State AI Governance Legislation Tracker, updated April 28, 2026, focuses specifically on cross-sectoral AI governance bills that apply to private sector organizations — not just government agencies — and identifies states including Colorado, California, Utah, Texas, New York, and Illinois as having enacted relevant frameworks.
What States Have Comprehensive AI Laws in Force or Imminent?
California — Highest State ASCSI™ Score (9/10)
California has more enacted AI laws than any other state — at least 20 by year-end 2025 — but two dominate compliance agendas.
SB 53 (Transparency in Frontier AI Act, TFAIA), signed by Governor Newsom on September 29, 2025, and effective January 1, 2026, made California the first state in the US to enact a statute specifically addressing frontier AI model safety. It requires “large frontier developers” — organizations training foundation models using more than 10²⁶ floating-point operations — to write, implement, and publicly publish a frontier AI safety framework; submit confidential catastrophic-risk assessments to the California Office of Emergency Services; and report critical safety incidents within 15 days. The California Attorney General enforces it with civil penalties up to $1,000,000 per violation.
AB 2013 (AI Training Data Transparency Act), also effective January 1, 2026, requires developers of generative AI systems available to Californians to post documentation of training data provenance on their public websites.
Several additional bills — including the California AI Transparency Act on digital content provenance — are phasing in through August 2026.
Sarah Mitchell notes the pattern: California’s laws address two different layers of the AI stack without trying to unify them. TFAIA hits the compute tier — who is building the largest models and whether they have thought through catastrophic risk. AB 2013 hits the data tier — what went into the model. Together, they create a traceability obligation from training inputs to deployment safety, which is a more coherent regulatory architecture than it first appears. Whether enforcement tracks the statute’s ambition is the question 2026 and 2027 will answer.
Colorado — Complete Rebuild After the Original Act Failed
Colorado made history in 2024 as the first state to pass a comprehensive AI law. Then it rewrote that law before it ever took effect.
The original Colorado AI Act (SB 24-205) required risk management programs, impact assessments, and a duty of care to prevent algorithmic discrimination across employment, housing, lending, insurance, healthcare, and education. Governor Polis signed it in May 2024 while publicly asking lawmakers to revisit it. After repeated legislative standoffs and a federal injunction filed by xAI in April 2026, the legislature adopted a new framework.
SB 26-189, signed May 14, 2026 and effective January 1, 2027, repeals and replaces SB 24-205. The new law drops risk management programs, impact assessments, and the duty-of-care obligation. In their place, it targets “covered automated decision-making technology” (ADMT) used in consequential decisions — employment, education, housing, financial services, insurance, healthcare, and government — and imposes four operational duties: advance notice before AI use, a post-adverse-outcome disclosure within 30 days, the right to correct inaccurate personal data, and meaningful human review and reconsideration upon request.
Scope note: the law expressly defines “consumer” to include employees and Colorado resident job applicants — reaching workforce decisions the Colorado Privacy Act largely excludes. Vendor contracts that indemnify a developer or deployer against liability for its own discriminatory ADMT use are void as against public policy.
Texas — Intent-Based Liability, Highest Penalty Ceiling
TRAIGA (Texas Responsible Artificial Intelligence Governance Act, HB 149), signed June 22, 2025, effective January 1, 2026, applies to any entity conducting business in Texas, offering products or services to Texas residents, or developing or deploying AI systems within the state. The Texas Attorney General’s consumer rights page confirms enforcement parameters.
TRAIGA prohibits AI systems designed to incite self-harm or violence, conduct social scoring by government entities, capture biometric data without consent, facilitate unlawful discrimination, or produce certain sexually explicit content. It requires healthcare AI disclosure before or at the time of service.
Texas’s primary innovation is intent-based liability: disparate impact alone is explicitly insufficient to demonstrate a violation under TRAIGA §552.056(c). An enforcement action requires proof of intent to cause harm or discriminate — a materially higher evidentiary bar than comparable laws in Colorado and California.
Civil penalties: $10,000–$12,000 per curable violation; $80,000–$200,000 per uncurable violation; $2,000–$40,000 per day for continuing violations. A 60-day cure period applies. The Texas AG holds exclusive enforcement authority; no private right of action exists. TRAIGA also preempts all city and county AI ordinances statewide, creating a single compliance standard.
New York — RAISE Act for Frontier AI, Effective January 2027
New York Governor Kathy Hochul signed the RAISE Act (Responsible AI Safety and Education Act) on December 19, 2025, and the final chapter amendment on March 27, 2026. As Davis Polk reported, the law takes effect January 1, 2027.
RAISE covers “frontier developers” who have trained a model using 10²⁶+ FLOPs; the “large frontier developer” category — with annual revenue exceeding $500 million — carries the most significant obligations, including publishing detailed safety and security frameworks, reporting critical safety incidents to the New York Department of Financial Services (NYDFS), and adhering to ongoing governance requirements. The NYDFS is empowered to administer and enforce, and the Attorney General may bring civil actions for violations.
New York sits alongside California as one of two states with explicit frontier-model safety legislation — a bicoastal bracket around the most capable AI systems built or deployed in the US.
Illinois — Employment AI, First Comprehensive Hiring Prohibition
Illinois has addressed AI in employment more extensively than any other state. Two laws matter:
AIVII (Artificial Intelligence Video Interview Act), effective January 1, 2020, requires employers to notify candidates before using AI to evaluate video interviews, obtain consent, explain how the AI works, and delete recordings upon request.
HB 3773 (Public Act 103-0804), signed August 9, 2024, effective January 1, 2026, amends the Illinois Human Rights Act to expressly prohibit employers from using AI that “has the effect of subjecting employees to unlawful discrimination” based on protected characteristics. The amendment covers employers, employment agencies, and labor organizations using AI for hiring, discipline, or promotion decisions. Enforcement flows through the Illinois Human Rights Act’s existing framework; the Illinois Department of Human Rights investigates complaints.
Utah — The Disclosure Baseline
SB 149 (Artificial Intelligence Policy Act), in force since 2024 with 2025 amendments, applies to regulated entities in consumer-facing interactions. The law requires disclosure when generative AI is used in consumer transactions upon the consumer’s request — a lower bar than California’s automatic disclosure requirements. The 2025 amendments clarified applicability to regulated entities and added mental health chatbot protections, requiring operators to maintain protocols preventing users from exposure to suicidal ideation content.
New York City — Strictest Employment AI Law in the US (ASCSI™ 9/10)
NYC Local Law 144 has required employers to conduct annual third-party bias audits of automated employment decision tools (AEDTs) used in hiring or promotion decisions for NYC roles, post bias audit results publicly, and notify candidates before AEDT use since July 5, 2023. The Department of Consumer and Worker Protection enforces it; penalties start at $500 per violation for first offenses, rising to $1,500 for subsequent violations. Candidates may request an alternative assessment if they object to the AEDT.
Which AI Laws Are Currently in Force vs. Coming in 2027?
| Jurisdiction | Law | In Force | Next Date |
|---|---|---|---|
| California | SB 53 (TFAIA) | Jan 1, 2026 | Annual transparency reports due |
| California | AB 2013 | Jan 1, 2026 | Ongoing |
| California | AI Transparency Act | Aug 2026 | Full obligations phase-in |
| Texas | TRAIGA (HB 149) | Jan 1, 2026 | AG rulemaking ongoing |
| Illinois | HB 3773 (AI employment) | Jan 1, 2026 | Ongoing |
| Utah | AI Policy Act (amended) | 2024, amended 2025 | Ongoing |
| Illinois | AIVII | Jan 1, 2020 | Ongoing |
| NYC | Local Law 144 | Jul 5, 2023 | Annual audit cycle |
| Colorado | SB 26-189 (new ADMT law) | May 14, 2026 signed | Jan 1, 2027 obligations |
| New York | RAISE Act (amended) | Mar 27, 2026 signed | Jan 1, 2027 obligations |
Source: Colorado General Assembly; California Legislative Information; Texas AG; NCSL; Davis Polk, May 2026. Retrieved July 30, 2026.
How Do State AI Laws Compare by Regulatory Approach?
State AI laws in 2026 cluster into three observable models, each reflecting different assumptions about where AI risk originates.
The frontier model model (California, New York): These states target the compute tier — the largest, most capable AI systems — and impose safety framework and incident reporting obligations on developers before harm materializes. The rationale is that catastrophic risk from foundation models warrants upstream intervention.
The deployment/decision model (Colorado, New York City): These laws target the moment AI influences a consequential decision about a real person. Colorado’s SB 26-189 and NYC’s Local Law 144 don’t regulate what models get built; they regulate what companies do with them when the output affects employment, housing, or credit. The focus is transparency to the affected individual and human recourse.
The intent-prohibition model (Texas, Illinois): TRAIGA and Illinois HB 3773 don’t require process — they prohibit outcomes. Texas makes intentional misuse illegal; Illinois makes AI-driven employment discrimination illegal under the Human Rights Act framework. Both states accept existing legal infrastructure and extend it to AI contexts rather than building new regulatory architecture.
The practical consequence for multistate compliance: a company operating in California, Colorado, Texas, New York, and Illinois simultaneously faces five distinct obligation sets that address different questions. California asks: is your frontier model safe? Colorado asks: did you tell users you’re using AI before making a consequential decision? Texas asks: did you intend to cause harm? Illinois asks: did your AI discriminate? New York asks: do you have a bias audit for your hiring tool? No single compliance program answers all five.
What Is the Axis State AI Compliance Stringency Index (ASCSI™)?
The ASCSI™ is an original Axis Intelligence Research metric that measures the regulatory compliance burden a company faces in each US state with enacted AI legislation, as of July 2026. It does not measure regulatory quality or consumer protection effectiveness — only the stringency of private-sector compliance obligations.
Methodology — Five dimensions, each scored 0–2, maximum 10 points:
| Dimension | 0 | 1 | 2 |
|---|---|---|---|
| Scope breadth | No law | Narrow (one sector or use case) | Comprehensive (multiple sectors or domains) |
| Enforcement mechanism | None | AG only, limited penalties | AG + penalties ≥ $100K or additional enforcer |
| Obligations stringency | None | Disclosure or notice only | Impact assessment, risk program, or audit required |
| Active status (as of July 2026) | Not yet signed | Signed, effective 2027 | In force |
| Consumer/individual rights | None | Limited (disclosure upon request) | Full (opt-out, appeal, human review) |
Formula (Axis Intelligence Research, baseline July 2026):
ASCSI™ = Scope(0–2) + Enforcement(0–2) + Obligations(0–2) + ActiveStatus(0–2) + ConsumerRights(0–2)
Scores — July 2026 baseline reading:
| Jurisdiction | Scope | Enforcement | Obligations | Active Status | Consumer Rights | ASCSI™ |
|---|---|---|---|---|---|---|
| California | 2 | 2 | 2 | 2 | 1 | 9/10 |
| New York City | 1 | 2 | 2 | 2 | 2 | 9/10 |
| Colorado | 2 | 1 | 1 | 1 | 2 | 7/10 |
| Texas | 1 | 2 | 1 | 2 | 1 | 7/10 |
| New York | 1 | 2 | 2 | 1 | 1 | 7/10 |
| Illinois | 1 | 1 | 1 | 2 | 1 | 6/10 |
| Utah | 1 | 1 | 1 | 2 | 1 | 6/10 |
| Tennessee | 1 | 1 | 1 | 2 | 1 | 6/10 |
Key observations from the ASCSI™ July 2026 baseline:
California and New York City share the top score (9/10) but through entirely different mechanisms. California reaches 9 via scope breadth — it covers frontier models, ADMT, training data, and deepfakes simultaneously. NYC reaches 9 via a narrow scope (hiring only) paired with the strictest obligation (mandatory third-party audits) and the strongest consumer rights (opt-out, public audit results). A company operating in both faces different compliance regimes that happen to score equally.
Colorado scores the same as Texas (both 7/10) despite opposite regulatory philosophies: Colorado prioritizes individual consumer rights and broad scope; Texas prioritizes enforcement teeth (the highest penalty ceiling of any US state AI law) and intent-based prohibitions.
Limitations of ASCSI™ at baseline: the score reflects enacted text, not enforcement intensity or regulatory guidance maturity. Colorado’s January 2027 ActiveStatus dimension (score: 1) will increase to 2 when obligations commence, raising its score to 8/10 — making it the second-highest-scoring state at that point. The metric will be updated at each quarterly reading.
Attribution: Axis Intelligence Research, ASCSI™ (State AI Compliance Stringency Index), baseline reading July 30, 2026. Licensed CC BY 4.0. To cite: “Axis Intelligence Research ASCSI™ Q3 2026, available at axis-intelligence.com.”
What Is the Federal vs. State AI Law Situation?
The most important structural fact about US AI regulation in 2026: there is no federal AI law. The closest things to binding federal AI standards are sector-specific guidance from the FTC, EEOC, and HHS — none of which are statutes — and the NIST AI Risk Management Framework, which is voluntary.
The Trump administration made two attempts to suppress state regulation.
First, Executive Order 14365, signed December 11, 2025, established an AI Litigation Task Force within the DOJ with a mandate to challenge state AI laws on Commerce Clause, preemption, and constitutionality grounds, and conditioned broadband BEAD funding on states’ willingness to pause AI statutes. Multiple legal observers noted that executive orders cannot independently preempt state law — only Congress can do that.
Second, the One Big Beautiful Bill Act included a 10-year moratorium on state and local AI regulation when the House passed it on May 22, 2025. The Senate stripped the provision by a 99–1 vote on July 1, 2025, according to Global Policy Watch. The Bill was signed into law on July 4, 2025 — without the moratorium. As MultiState analyzed, the practical impact: “states are free to enforce current and upcoming state AI laws.”
The bipartisan nature of the moratorium’s rejection is the more interesting data point. Senator Marsha Blackburn (R-TN), who withdrew her support and introduced the stripping amendment, cited drafting flaws and overreach — concerns voiced across the ideological spectrum from Cantwell (D-WA) to Greene (R-GA). What united the critics was the practical reality that states had already moved faster and more specifically than any federal framework was positioned to.
Sarah Mitchell: The federal-vs-state dynamic here is the inverse of what most people assume. The conventional expectation is that federal law preempts a messy state patchwork. What actually happened is that 50 state legislatures ran a distributed policy experiment — deepfake laws, frontier model requirements, employment AI prohibitions, bias audit mandates — and produced a body of law that is more granular and more varied than anything a single national standard would have generated. The moratorium’s defeat means that experiment continues. The compliance cost of navigating it is real; the policy learning from it is also real.
What Are the Most Common Categories of State AI Legislation?
Based on NCSL and MultiState tracking data through early 2026, six categories dominate enacted legislation:
Deepfake laws are the most numerously enacted category. More than 20 states enacted nonconsensual explicit deepfake laws in 2024 alone, per MultiState, with additional states adding election deepfake prohibitions in 2025. These laws are narrow and targeted — they do not create a compliance burden for organizations deploying enterprise AI.
Government use of AI generated approximately 250 legislative measures in 40+ states in 2025, per NCSL. Most require government agencies to publish AI system inventories, establish human oversight for high-stakes government AI, or ban certain uses (Kansas prohibited AI platforms from foreign-adversary countries on state-issued devices; Montana’s Right to Compute Act affirmed residents’ right to use AI without government interference).
AI in healthcare reached at least 10 enacted state laws in 2025. Arizona (HB 2175) prohibited using AI to deny medical claims. Illinois (HB 1806) banned AI use in therapy or psychotherapy services. Texas (SB 1188) required medical providers to review and approve AI-generated records before acting on them.
AI in hiring and employment decisions is the fastest-growing enterprise compliance category: Illinois HB 3773 (effective 2026), NYC Local Law 144 (in force since 2023), and Colorado SB 26-189 (effective 2027) all impose obligations on employer use of automated decision tools affecting workers and applicants.
Chatbot transparency and consumer interaction saw at least 5 states — California, Maine, New Hampshire, New York, and Utah — enact legislation in 2025 requiring disclosure when consumers interact with AI rather than a human, per NCSL.
AI training data and provenance laws, led by California’s AB 2013, require developers to document and disclose what data was used to train AI systems available to consumers.
How Is AI Regulation in Hiring Actually Enforced?
NYC Local Law 144 provides the most concrete enforcement data because it has been operating the longest.
The New York City Department of Consumer and Worker Protection (DCWP) enforces it. The law requires employers and employment agencies to: conduct an annual third-party bias audit of any automated employment decision tool (AEDT) used in hiring or promotion decisions; publish the most recent audit results publicly on their websites; notify candidates in job postings or when applying that an AEDT is used; and inform candidates which characteristics the AEDT evaluates.
Penalty structure: $500 for the first violation in a calendar day, $1,500 for each subsequent violation per calendar day. The relatively modest penalty ceiling has not stopped DCWP from issuing enforcement guidance and pursuing compliance actions. Candidates who object to AEDT evaluation must be offered an alternative assessment process if one reasonably exists.
Illinois HB 3773 (effective January 1, 2026) takes a different approach: it bans discriminatory outcomes from AI, treating AI-driven discrimination as a civil rights violation under the Illinois Human Rights Act. This puts the enforcement machinery of the IDHR behind it — the same agency that investigates employment discrimination more broadly. The discrimination standard, not a process audit standard, is what matters here.
Which State AI Laws Carry the Highest Penalties?
| Jurisdiction | Law | Maximum Penalty | Enforcer | Private Right of Action |
|---|---|---|---|---|
| California (SB 53) | TFAIA | $1,000,000 per violation | Attorney General | No |
| Texas (TRAIGA) | HB 149 | $200,000 per uncurable violation; $40,000/day continuing | Attorney General | No |
| Colorado (SB 26-189) | ADMT law | TBD via AG rulemaking | Attorney General | No |
| New York (RAISE Act) | RAISE | TBD via DFS rulemaking | DFS + Attorney General | No |
| Illinois (HB 3773) | Human Rights Act | Civil damages, attorney fees | IDHR | Yes — via Human Rights Act |
| NYC (Local Law 144) | AEDT bias audits | $1,500 per violation per day | DCWP | No |
| Utah | AI Policy Act | Civil penalties under UCSPA | Division of Consumer Protection | No |
Source: Texas AG; California Legislative Information; Ogletree Deakins; Davis Polk; DCWP; NCSL. Retrieved July 30, 2026.
Illinois is the outlier: it is the only jurisdiction listed above that provides a private right of action for affected individuals, via the existing Illinois Human Rights Act framework. Every other state AI law concentrates enforcement in a government agency.
What Does the State AI Regulatory Map Look Like by Region?
The regulatory intensity is geographically uneven in ways that matter for business planning.
Pacific Coast: California has the most laws and the highest ASCSI™ score. Washington, Oregon, and Hawaii have active bill pipelines but no comprehensive enacted law as of July 2026.
Northeast + Mid-Atlantic: New York leads in frontier model regulation (RAISE Act) and employment AI (NYC LL 144); Massachusetts, Connecticut, and New Jersey have active bills that have not yet cleared both chambers.
South / Sun Belt: Texas leads the region with TRAIGA, the only comprehensive AI governance law in the South or Southwest, with an unusual combination of high penalty ceilings and intent-based liability. Florida, Georgia, and North Carolina have introduced bills but not enacted comprehensive AI frameworks.
Midwest: Illinois leads with the AIVII (2020) and HB 3773 (2026). Michigan and Ohio have active legislative pipelines.
Mountain West / Interior: Colorado and Utah have enacted frameworks; both are refining them. Arizona enacted a meaningful health insurance AI bill (HB 2175). Montana enacted the Right to Compute Act, which cuts the other direction — affirming computational freedoms rather than imposing restrictions.
Axis Intelligence Research observes that the regulatory map roughly tracks economic geography: the states with the largest AI deployment footprints — California, Texas, New York — also have the most developed legal frameworks. That is not a coincidence. Regulatory attention follows commercial activity.
What Sectors Are Most Affected by State AI Laws?
State AI laws, read collectively, impose the densest obligation set on organizations using AI in five sectors:
Healthcare: Arizona’s claim-denial prohibition, Illinois’s therapy ban, Texas’s medical record review requirement, and California’s data provenance obligations collectively make healthcare AI the most heavily targeted sector in state law. A health system deploying AI for prior authorization, clinical decision support, or patient communication must track different state obligations for each state in which it operates.
Employment and hiring: NYC LL 144, Illinois HB 3773 and AIVII, and Colorado SB 26-189’s inclusion of employees and job applicants in its consumer definition create a de facto national standard for high-volume employers: conduct bias audits, disclose AI use, and offer recourse. These three jurisdictions alone cover roughly 35 million workers.
Financial services: Most comprehensive AI laws list credit, lending, and insurance as covered consequential decision domains. California, Colorado, and Texas all reach financial sector AI decisions. Utah’s law applies to regulated entities, including financial services firms.
Real estate and housing: Algorithmic rental pricing has attracted specific attention in New York and California (anti-price-fixing bills) and is a covered domain in Colorado’s ADMT law and California’s future automated decision framework.
Public-facing AI interactions (chatbots, companions, consumer AI): The emerging chatbot transparency category reached 5 enacted state laws in 2025, with more expected. Operators of companion AI, mental health chatbots, and customer service AI face increasing disclosure and safety obligations.
FAQ
What is the difference between a comprehensive AI law and a narrow AI law?
A comprehensive AI law, like California’s SB 53 or Colorado’s SB 26-189, establishes an obligation framework that applies across multiple sectors and use cases. It typically requires companies to take affirmative steps — disclosures, audits, risk assessments, or incident reporting — before or when deploying AI in covered contexts. A narrow AI law targets a specific use case: deepfakes, AI-generated election content, AI-conducted video interviews, or AI-denied insurance claims. Narrow laws often work by adding AI language to existing statutes (criminal codes, consumer protection laws, anti-discrimination laws) rather than creating new regulatory regimes.
Does the Trump executive order on AI preempt state AI laws?
No. Executive Order 14365 (December 11, 2025) directed the DOJ to establish an AI Litigation Task Force to challenge state AI laws in court and conditioned some federal broadband funding on states’ compliance with federal AI priorities. But executive orders cannot directly preempt state law — that requires an act of Congress. As White & Case noted, the order would “likely not independently displace state AI laws.” Organizations must continue complying with applicable state laws while monitoring federal litigation developments.
Which state has the strictest AI law for employers?
It depends on the enforcement theory. New York City has the strictest process requirements for employers — mandatory third-party bias audits, public results disclosure, and candidate opt-out rights under Local Law 144. Illinois has the broadest prohibition scope — AI that has the effect of discrimination is banned under HB 3773, regardless of intent. California’s AB 2013 and TFAIA don’t directly regulate employer AI decisions but impose developer obligations that flow down to AI vendors used by employers.
What is the ASCSI™ and how is it calculated?
The Axis State AI Compliance Stringency Index (ASCSI™) is an original Axis Intelligence Research metric — see Section 8 above for full methodology. It scores states on five dimensions (scope breadth, enforcement mechanism, obligations stringency, active status, and consumer/individual rights), each 0–2 points, for a maximum of 10. The July 2026 baseline reading places California and NYC jointly at the top (9/10). The metric will be updated quarterly as laws take effect or are revised.
Is there a US federal AI law?
Not as of July 2026. The NIST AI Risk Management Framework (published February 2023) is voluntary guidance, not binding law. The One Big Beautiful Bill Act, signed July 4, 2025, did not include the AI moratorium provision after the Senate stripped it 99–1. The Trump administration’s AI executive orders direct federal agency conduct and challenge state laws but do not create enforceable federal AI obligations on private companies.
What is the biggest compliance risk in 2027?
Colorado and New York both have laws taking effect January 1, 2027 — SB 26-189 and the RAISE Act, respectively. For companies using AI in consequential employment, housing, or financial decisions with Colorado consumers, SB 26-189’s advance notice, post-decision disclosure, data correction, and human review obligations represent a material process change. For frontier AI developers with $500M+ annual revenue, New York’s RAISE Act requirements — safety frameworks, incident reporting to the NYDFS, and DFS oversight — kick in simultaneously. The practical risk: two major compliance buildouts with the same go-live date and no federal coordination to simplify them.
How do US state AI laws compare to the EU AI Act?
The contrast is structural. The EU AI Act is a single, unified regulation covering all EU member states under a risk-based classification system, with high-risk AI obligations phased in from August 2, 2026. US state AI laws are 50 independent experiments with different scopes, different definitions, different enforcement agencies, and different penalty structures. Axis Intelligence Research’s ARCSI™ global score gives the EU 9.5/10 and the US federal level 3.4/10; the ASCSI™ state scores show individual states — California, NYC — approaching EU-level stringency in narrow domains. See the companion article AI Regulation by Country 2026 for the global comparison.
Which states are expected to pass new AI laws before year-end 2026?
Based on active bill pipelines tracked by MultiState and NCSL as of March 2026, the states with the most advanced AI governance bills in committee or on the floor include: Massachusetts (automated decision-making transparency), New Jersey (AI in employment), Washington (algorithmic accountability), Maryland (AI procurement and government use), and Pennsylvania (healthcare AI). None of these had enacted comprehensive private-sector AI obligations as of July 30, 2026.
Methodology
Data collection: Primary source materials were fetched during the production session of this article (July 30, 2026). Sources include the NCSL Artificial Intelligence Legislation Database, MultiState AI Legislation Tracker (2024, 2025, 2026 data), IAPP US State AI Governance Legislation Tracker, Colorado General Assembly bill text for SB 26-189, Texas Attorney General Consumer AI Rights page, California Legislative Information for SB 53 and AB 2013, Davis Polk analysis of the RAISE Act, and Wiley Law RAISE Act analysis.
ASCSI™ formula: All five dimension inputs and weights are disclosed in full in the dedicated section above. Scores for each dimension are determined by the factual status of enacted law as of July 2026. The formula is: ASCSI™ = Scope(0–2) + Enforcement(0–2) + Obligations(0–2) + ActiveStatus(0–2) + ConsumerRights(0–2). Maximum possible score: 10. Arithmetic verified in Python (July 30, 2026). The ASCSI™ is an Axis Intelligence Research original metric, trademark pending.
Legislative counts: Bill introduction and enactment totals come from MultiState (2023–2026 data) and NCSL (2025 enacted measures). Where sources differ in exact count, the more conservative figure is used and the discrepancy noted.
Scope of coverage: This article covers enacted state AI laws with private-sector applicability. It does not cover federal regulations, voluntary frameworks (NIST AI RMF), AI provisions embedded in general consumer protection enforcement guidance, or AI-adjacent laws (data privacy statutes not specifically targeting AI systems) unless those laws are directly amended by AI-specific legislation.
Limitations: The legislative landscape changes faster than any static article can track. The ASCSI™ scoring reflects text as enacted; actual enforcement intensity may differ. Colorado’s and New York’s scores will change when January 2027 obligations commence. This article is updated on a relevance-triggered basis — when a new state law is enacted or federal preemption legislation advances.
About This Dataset
The companion CSV (us-ai-laws-by-state.csv) contains every quantitative claim in this article as a row, with source organization, source document, URL, retrieval date, and Axis-calculated flags where applicable. The ASCSI™ component data appears with full methodology notation in the axis_calculated column.
Dataset license: CC BY 4.0. Attribution: “Axis Intelligence Research, US AI Laws by State Statistics 2026, axis-intelligence.com.”
To cite this article:
APA: Axis Intelligence Research & Mitchell, S. (2026). US AI laws by state 2026: The complete compliance map. Axis Intelligence Research. https://axis-intelligence.com/us-ai-laws-by-state/
MLA: Axis Intelligence Research and Sarah Mitchell. “US AI Laws by State 2026: The Complete Compliance Map.” Axis Intelligence Research, 30 July 2026, axis-intelligence.com/us-ai-laws-by-state/.
Chicago: Axis Intelligence Research and Sarah Mitchell. “US AI Laws by State 2026: The Complete Compliance Map.” Axis Intelligence Research. July 30, 2026. https://axis-intelligence.com/us-ai-laws-by-state/.