State AI laws in 2027: who they reach and what deployers owe
Colorado's rewritten act and California's automated decision rules both bite on January 1, 2027; Texas, Illinois, Utah and New York add disclosure, employment and public-sector duties, and federal preemption is still a policy aim.
Six states, three kinds of duty
A bank, health system or public agency that uses AI in decisions about people will meet state law in 2027 through three routes. Two states regulate the decision itself: Colorado through its rewritten automated decision-making technology act, which applies to consequential decisions made on or after January 1, 2027 [2], and California through privacy rules that require businesses already using automated decisionmaking technology for significant decisions to comply by the same date [10]. Texas and Utah regulate conduct and disclosure: Texas has applied its AI governance act since January 1, 2026 [6], and Utah requires licensed professionals to disclose generative AI in high-risk interactions [14]. Illinois, California and New York add rules for AI in employment, including in the public sector [13] [12] [17].
Above them sits a federal policy that wants fewer state AI laws but has not displaced any. Executive Order 14365 sets up a Justice Department task force to challenge state AI laws and asks for a federal framework that would preempt those that conflict with its policy [21]. Statuses below are as checked on November 16, 2026.
Moraph's view
Moraph's view: the useful unit for 2027 planning is the decision, not the model. Each of these laws attaches duties to a use (a lending decision, a hiring screen, a chatbot talking to a patient or a resident), so one system can fall under several laws at once and a second system under none. An inventory of AI uses, tagged by decision type and state, is the starting point for all of them.
Colorado: a rewritten act with a fixed date
Colorado enacted Senate Bill 24-205 in 2024, and a 2025 special session bill, SB25B-004, signed on August 28, 2025, moved its requirements to June 30, 2026 [3]. Before that date arrived, SB26-189, signed on May 14, 2026, repealed and reenacted the framework as rules on automated decision-making technology in consequential decisions [1]. The act takes effect January 1, 2027 and applies to consequential decisions made on or after that date; the rulemaking provisions took effect on passage [2].
Coverage turns on whether a technology materially influences a consequential decision, meaning its output is a non-de minimis factor that affects the outcome, including by ranking, scoring, recommending or classifying [2]. The covered domains are education, employment, the lease or purchase of residential real estate in Colorado, financial or lending services, insurance, health-care services, and essential government services and public benefits [2]. The act leaves out technologies used for cybersecurity, anti-money laundering, sanctions compliance and fraud prevention, and tools used only to summarize, organize or present information for human review that produce no score, ranking or other inference that materially influences the outcome [2]. Employees, and job applicants who are Colorado residents, count as consumers [2].
A deployer must give a clear and conspicuous notice before it uses a covered technology to materially influence a consequential decision, and a prominent public notice at points of consumer interaction satisfies that duty [2]. After an adverse outcome, it must provide within 30 days a plain language description of the decision and the technology's role, a process to request more information about the technology and its inputs, and an explanation of the consumer's rights [2]. A consumer with an adverse outcome may then ask for instructions for correcting inaccurate personal data, and for meaningful human review and reconsideration of the decision to the extent commercially reasonable [2]. Deployers must keep records reasonably necessary to show compliance for at least 3 years after each decision [2]. A creditor whose ECOA and Regulation B notice also meets the act's requirements need not send a separate one [2].
Health systems are mostly outside the core duties. Sections 6-1-1701 to 6-1-1706 do not apply to HIPAA covered entities or their business associates for services rendered to them, except for decisions about employment, and for a health-care provider that exemption applies only if it operates from a location within Colorado [2]. Covered entities must still give patients a general notice of their use of advanced technologies, and one that uses a covered technology to decide eligibility for financial assistance must give the patient a description of the decision, the information relied on, and how to request correction and human review [2]. Insurers subject to Colorado's existing insurance statute on algorithms and predictive models, section 10-3-1104.9, are in compliance in the practice of insurance [2].
Public agencies are a harder read. The act defines a deployer as a person doing business in Colorado that deploys a covered technology, and it lists essential government services and public benefits as a covered domain [2]; the text of the new part does not say in terms whether a state or local agency is a deployer [2].
The Attorney General enforces the act as a deceptive trade practice, it creates no new private right of action, and the Attorney General must give a notice of violation and 60 days to cure where a cure is possible, a requirement repealed on January 1, 2030 [2]. The act requires rules on the post-adverse outcome disclosures and on the consumer rights section on or before January 1, 2027 [2]. The Department of Law filed proposed rules with the Secretary of State on August 11, 2026, released an interim draft on October 6, 2026 and held its rulemaking hearing on October 26, 2026 [4]. Enforcement is also paused by a federal court. In X.AI LLC v. Weiser, a challenge to the 2024 law that the Justice Department has joined, a magistrate judge ordered on April 27, 2026 that the Attorney General not initiate enforcement, including investigations, of SB24-205 or any replacement enacted in the 2026 session for violations that occur on or before 14 days after the court rules on xAI's motion for a preliminary injunction [23] [24]. Under the same order, xAI must file that motion within 28 days after final rules are adopted [23]. A separate law, HB26-1263, signed on May 29, 2026, sets duties for operators of conversational AI services available to the public, including disclosing that the service is AI, with requirements effective January 1, 2027 [5].
Moraph's view
Moraph's view: the new Colorado text is a notice-and-redress law. As we read Part 17, it contains no impact assessment or risk management program duty for deployers; the work is knowing which decisions a tool materially influences, posting the notice, producing the 30-day explanation, and staffing a human review that the act defines with some care. "Materially influence" is a lower bar than California's "substantially replace human decisionmaking", so a tool that a person reviews can be outside California's rules and inside Colorado's. The final rules will decide how heavy the adverse outcome notice is. The January 1, 2027 date is fixed in statute, but because of the court-ordered pause, the first Colorado enforcement action cannot come until after the federal court rules on xAI's motion.
California: automated decision rules in privacy law, and in employment law
The California Privacy Protection Agency Board adopted regulations on July 24, 2025 that, among other things, implement consumers' rights to access and opt out of businesses' use of automated decisionmaking technology; the Office of Administrative Law approved them on September 22, 2025, and they took effect January 1, 2026 [9]. The rules define such technology as technology that processes personal information and uses computation to replace or substantially replace human decisionmaking, and a human only counts as involved if the reviewer knows how to interpret the output, reviews it with other relevant information and has authority to change the decision [10].
A significant decision is one that results in the provision or denial of financial or lending services, housing, education enrollment or opportunities, employment or independent contracting opportunities or compensation, or healthcare services, and financial or lending services include credit, funds transfers and deposit accounts [10]. A business already using such technology for a significant decision must comply by January 1, 2027 [10]. It must give a pre-use notice that states the specific purpose, offer an opt-out unless an exception applies, such as an appeal to a human reviewer with authority to overturn the decision, and answer access requests [10]. Using the technology for a significant decision also requires a risk assessment: for processing begun before the rules took effect and still continuing, it must be completed by December 31, 2027, and information on assessments conducted in 2026 and 2027 is due to the Agency by April 1, 2028 [10].
The privacy statute itself narrows the reach for banks and health systems: it does not apply to personal information collected or processed subject to the Gramm-Leach-Bliley Act, or to protected health information collected by a HIPAA covered entity or business associate [11]. Employment is covered separately. The Civil Rights Council's employment regulations on automated-decision systems under the Fair Employment and Housing Act were filed with the Secretary of State on June 27, 2025 and took effect October 1, 2025 [12].
Moraph's view
Moraph's view: for a bank, the Gramm-Leach-Bliley carve-out takes much customer lending data outside the CCPA rules, but hiring, promotion and pay decisions about California staff are reached by the Civil Rights Council rules regardless. For a health system the same pattern holds: patient data under HIPAA sits outside, workforce decisions do not. Banks and hospitals should expect their first California AI compliance questions to come from human resources, not from customer channels.
Texas: prohibitions for everyone, disclosure for agencies and clinicians
The Texas Responsible Artificial Intelligence Governance Act, HB 149, was signed on June 22, 2025 and took effect January 1, 2026 [6]. It applies to anyone who conducts business in Texas, produces a product or service used by Texas residents, or develops or deploys an AI system in the state [7].
Its disclosure duties fall on government and health care. A governmental agency that makes available an AI system intended to interact with consumers must disclose, before or at the time of interaction, that the consumer is interacting with AI, written plainly and without dark patterns, even when it would be obvious [7]. A governmental entity may not use AI for social scoring, or to identify a specific individual from biometric data gathered without consent where that would infringe the individual's rights; hospital districts and institutions of higher education are outside the definition of governmental entity [7]. When AI is used in relation to a health care service or treatment, the provider must disclose it to the patient or the patient's representative by the date the service is first provided, or as soon as reasonably possible in an emergency [7]. The act also requires the Department of Information Resources to collect from each state agency an evaluation of its use or considered use of AI systems [7].
For private deployers the act works through prohibitions. No one may develop or deploy AI with the intent to unlawfully discriminate against a protected class, and a disparate impact alone does not show that intent [7]. A federally insured financial institution is considered to comply with that section if it complies with federal and state banking laws, and insurance entities subject to the insurance code's unfair discrimination rules are outside it [7].
The Attorney General has exclusive enforcement authority, there is no private right of action, and the Attorney General must give 60 days to cure before suing [7]. Uncured violations carry civil penalties of $10,000 to $12,000 for a curable violation, $80,000 to $200,000 for an uncurable one, and $2,000 to $40,000 for each day a violation continues [7]. A defendant is not liable if it finds a violation through testing or through an internal review while substantially complying with the NIST generative AI profile or another recognized AI risk framework [7]. The Attorney General's office now runs an online AI complaint form and a page describing the act [8].
Moraph's view
Moraph's view: TRAIGA asks little of a bank's lending models, because the discrimination section requires intent and gives insured banks a compliance presumption. Its practical reach is the front door: every Texas agency chatbot and every clinic that uses AI in care needs a disclosure, and the Attorney General's complaint form gives residents a direct route to report a missing one.
Employment and public-sector rules: Illinois, Utah and New York
Illinois Public Act 103-0804 amended the Human Rights Act, effective January 1, 2026, to make it a civil rights violation for an employer to use AI that has the effect of subjecting employees to discrimination in recruitment, hiring, promotion, discipline, discharge or the terms of employment, or to use zip codes as a proxy for protected classes [13]. Failing to notify employees that AI is being used for those purposes is also a violation, and the Department is to adopt rules on when and how notice is given [13]. The Act's definition of employer includes the State and any political subdivision, municipal corporation or other governmental unit regardless of the number of employees [13].
Utah's S.B. 226 of 2025 requires anyone providing services in an occupation that the Department of Commerce licenses or certifies to disclose prominently when a client is interacting with generative AI in a high-risk interaction, which covers collecting health, financial or biometric data and giving financial, legal, medical or mental health advice [14]. The disclosure is given orally at the start of a spoken interaction and in writing before a written one; any supplier using generative AI in a consumer transaction must also disclose it if a person clearly asks, and the Division of Consumer Protection may fine up to $2,500 per violation [14]. The state's Artificial Intelligence Policy Act was originally due for repeal on May 1, 2025; two 2025 bills moved that date to July 1, 2027 [14] [15], and the Utah Code still showed July 1, 2027 when checked [16].
New York's State Technology Law now has an article on automated employment decision-making in state government [17]. A state agency that uses software built on algorithms, computational models or AI to automate or replace human employment decisions, such as pay, hiring, discipline, promotion or termination, must publish a list of those tools on its website and update it each year [18] [19]. Counties, cities, towns, villages, school districts, the State University of New York, the City University of New York and community colleges must publish similar lists to the extent feasible [18] [20]. The definitions section and the local list requirement carry a repeal date of July 1, 2028, and the state agency section has a new version from that date [18] [19] [20].
Moraph's view
Moraph's view: employment is where state AI law reaches the most deployers in 2027. Every bank, health system and agency is an employer, and the Illinois and California employment rules apply to them whatever their sector exemptions say elsewhere. The New York lists are modest duties, but they make public agencies' hiring and discipline tools visible in a way no other state has required.
Federal preemption: a policy aim, not yet a statute
Executive Order 14365 of December 11, 2025 told the Attorney General to establish an AI Litigation Task Force within 30 days to challenge state AI laws that conflict with its policy, told the Secretary of Commerce to publish within 90 days an evaluation identifying onerous state laws, and tied remaining BEAD non-deployment funds to that evaluation [21]. It names Colorado's law banning algorithmic discrimination as an example of the state rules it opposes [21], and on April 24, 2026 the Justice Department moved to intervene in xAI's lawsuit against that law, arguing that it violates the Equal Protection Clause [24]. The order also says that the legislative recommendation it calls for shall not propose preempting otherwise lawful state laws on state government procurement and use of AI [21].
The White House's March 2026 legislative recommendations ask Congress to preempt state AI laws that impose undue burdens, but not states' laws of general applicability against AI developers and users, including laws to prevent fraud and protect consumers, and not requirements governing a state's own use of AI [22].
Moraph's view
Moraph's view: as checked on November 16, 2026 we found no federal statute that preempts these state laws, and we did not locate the Commerce evaluation on an official Commerce page. The administration's own carve-outs matter for public agencies: the Texas, New York and Illinois duties on government use of AI are the rules least likely to be preempted. For private deployers, we read the Colorado court order as limiting when the Attorney General may enforce, not when the duties begin, and the California date stands unless a court or Congress acts.
The 2027 map at a glance
| Law | Who it reaches | What deployers must do | Date, as checked on November 16, 2026 |
|---|---|---|---|
| Colorado SB26-189 (automated decision-making technology) [1] | Developers and deployers doing business in Colorado; HIPAA covered entities exempt except employment decisions [2] | Notice, adverse outcome disclosure within 30 days, correction and human review, records kept 3 years [2] | Effective January 1, 2027; rules due the same day [2]; enforcement paused by court order [23] |
| Colorado HB26-1263 (conversational AI services) [5] | Operators of conversational AI services available to the public [5] | AI disclosure, age estimation, self-harm protocol, annual report [5] | Requirements effective January 1, 2027 [5] |
| California CCPA rules on automated decisionmaking technology [9] | Businesses under the CCPA; GLBA data and HIPAA PHI outside the statute [11] | Pre-use notice, opt-out or human appeal, access, risk assessment [10] | Comply by January 1, 2027; risk assessments by December 31, 2027 [10] |
| California FEHA rules on automated-decision systems [12] | Employers under the Fair Employment and Housing Act | Employment regulations on automated-decision systems | Effective October 1, 2025 [12] |
| Texas HB 149 (TRAIGA) [6] | Anyone doing business or deploying AI in Texas [7] | Agency and health care AI disclosure; bans on social scoring and intentional discrimination [7] | Effective January 1, 2026 [6] |
| Illinois Public Act 103-0804 [13] | Employers, including state and local government of any size [13] | No discriminatory AI use or zip code proxies; notice to employees [13] | Effective January 1, 2026 [13] |
| Utah S.B. 226 [14] | Licensed or certified occupations; suppliers in consumer transactions [14] | Generative AI disclosure in high-risk interactions, and on request [14] | Disclosure duties effective May 7, 2025 [14]; AI Policy Act repeal date July 1, 2027 [16] |
| New York State Technology Law, Article 4 [17] | State agencies; local governments, school districts, SUNY and CUNY [18] | Annual public list of automated employment decision-making tools [19] [20] | Definitions and local list duty repealed July 1, 2028 [18] [20] |
What to watch through 2027
- The Colorado Attorney General's final rules on adverse outcome disclosures and consumer rights, due on or before January 1, 2027 [2], after a hearing held on October 26, 2026 [4].
- Whether the 2027 Colorado session reopens the act a third time, after the 2025 delay and the 2026 rewrite [3] [1].
- Utah's July 1, 2027 repeal date for the Artificial Intelligence Policy Act [16].
- The first California risk assessments for automated decisions, due by December 31, 2027 for processing begun before the rules took effect [10].
- Whether the Texas Attorney General brings a case under the act [7], and how the federal court rules on xAI's motion against Colorado's law, which the Justice Department has joined [23] [24].
Scope and method
We read the enacted text, official code or agency rulemaking record for each law in this piece on November 16, 2026 and took dates from the legislature's or the agency's own page. The piece covers state laws that put duties on organizations using AI in decisions about people or in dealings with the public. It leaves out laws aimed only at developers of large models, insurance regulators' AI bulletins (covered in our financial services outlook), and state rules specific to clinical AI and prior authorization (covered in our health care outlook). This is a description of the laws, not legal advice on how they apply to a given organization.
What this means for you
- Banks: Map lending, deposit and hiring decisions that a model materially influences to Colorado's notice and 30-day adverse outcome duties; a Regulation B notice that also meets the act's content covers it, and AML, sanctions and fraud tools are outside [2]. In California, check which decision data falls outside the GLBA carve-out [11], and treat hiring tools as in scope of the employment rules [12]. In Texas, insured banks that comply with banking law are deemed to comply with the discrimination section [7].
- Health systems: In Colorado, the HIPAA exemption does not cover employment decisions, and patients must get a general notice of advanced technology use plus specific disclosures where a tool decides financial assistance eligibility [2]. In Texas, tell patients when AI is used in their care by the date care is first provided [7]; in Utah, licensed staff must disclose generative AI in high-risk interactions [14].
- Public agencies: Texas agencies need an AI disclosure on every consumer-facing AI system and must include an evaluation of their use or considered use of AI systems in the information the Department of Information Resources collects [7]. New York agencies must publish and update their lists of automated employment decision-making tools [19], and Illinois public employers owe employees notice of AI in employment decisions [13].
- AI vendors: From January 1, 2027, developers selling into Colorado must give deployers documentation of intended uses, training data categories, limitations and human review instructions, and contract terms that shift liability for a party's own discriminatory acts are void [2]. The Texas Attorney General may demand system descriptions, metrics and monitoring details after a complaint [7].
Moraph predictions
Every prediction is Moraph's view, not a fact. Each has a date, a confidence level (70% means we expect to be wrong about three times in ten) and a test anyone can apply. We score them on this page when the date comes and never remove one. How we make and score predictions.
P1 By Confidence 75% Open
The Colorado General Assembly does not enact a bill in its 2027 regular session that delays the January 1, 2027 effective date of Part 17 (automated decision-making technology) or repeals the deployer disclosure and consumer rights duties in sections 6-1-1704 and 6-1-1705.
How we will check: Search the Colorado General Assembly's 2027 session laws on leg.colorado.gov for any act amending part 17 of article 1 of title 6; the call is right if none signed on or before July 31, 2027 postpones the act or repeals sections 6-1-1704 or 6-1-1705.
Basis: SB26-189 repealed and reenacted the 2024 act and set January 1, 2027 for decisions made from that date [2] [1], after SB25B-004 delayed the original law once [3], and the Attorney General's rulemaking reached its hearing on October 26, 2026 [4].
P2 By Confidence 70% Open
Utah changes the July 1, 2027 repeal date of its Artificial Intelligence Policy Act (Title 13, Chapter 72), extending it or removing it, before that date passes.
How we will check: On July 1, 2027, Utah Code section 63I-2-213 on le.utah.gov shows a repeal date for Title 13, Chapter 72 later than July 1, 2027, or no longer lists the chapter; the call is wrong if the chapter is repealed on July 1, 2027.
Basis: The act's original repeal date of May 1, 2025 was extended to July 1, 2027 by two 2025 bills [14] [15], and the current code shows that date [16].
P3 By Confidence 55% Open
The Texas Attorney General files at least one lawsuit that pleads a violation of the Texas Responsible Artificial Intelligence Governance Act (Business & Commerce Code chapter 552).
How we will check: A press release or court petition published by the Office of the Texas Attorney General, dated on or before December 31, 2027, that cites chapter 552 of the Business & Commerce Code or TRAIGA as a basis for the claims.
Basis: The act has applied since January 1, 2026 [6]; the Attorney General has exclusive enforcement authority, must allow 60 days to cure, and may use civil investigative demands after complaints [7], and the office runs an online AI complaint form [8].
P4 By Confidence 65% Open
No court grants a preliminary or permanent injunction against enforcement of any provision of Colorado's automated decision-making technology act (sections 6-1-1701 to 6-1-1709) by December 31, 2027; the stay of enforcement entered by agreement on April 27, 2026 in X.AI LLC v. Weiser does not count.
How we will check: As of December 31, 2027, the docket of X.AI LLC v. Weiser (D. Colo. No. 1:26-cv-01515), and any other federal or Colorado case naming the Colorado Attorney General and Part 17, shows no order granting a preliminary or permanent injunction against enforcement of Part 17; the call is wrong if such an order is entered on or before that date, even if later reversed.
Basis: xAI must move for a preliminary injunction within 28 days after final rules are adopted [23], the Justice Department has joined the suit [24], and Executive Order 14365 names Colorado's law [21]; but the act as reenacted is a disclosure and consumer rights law with no private right of action [2], and the White House's own recommendations would leave generally applicable consumer protection laws with the states [22].
P5 By Confidence 70% Open
The California Privacy Protection Agency does not announce an enforcement action that alleges a violation of its automated decisionmaking technology rules (Article 11 of the CCPA regulations) before the end of 2027.
How we will check: No press release, Board decision or stipulated order published on cppa.ca.gov on or before December 31, 2027 alleges a violation of sections 7200 to 7222 of title 11 of the California Code of Regulations.
Basis: Businesses already using such technology must comply only from January 1, 2027, and risk assessments for processing begun before the rules took effect are due by December 31, 2027 with submissions by April 1, 2028 [10]; the rules took effect January 1, 2026 [9].
Facts on this page link to their sources. Text marked "Moraph's view", and every prediction, is our own analysis, not a fact or advice. Methodology.
Spotted an error? Tell us and we will check it and correct the piece where needed.