GuideTexas HB 149 · TRAIGAVersion 1.025 October 2026
Texas HB 149, the Texas Responsible AI Governance Act
Texas HB 149, the Texas Responsible Artificial Intelligence Governance Act, signed on 22 June 2025 and in effect since 1 January 2026, is enforced by the attorney general alone after a 60-day notice to cure, and protects a company that finds a violation through internal review while substantially complying with a recognized AI risk management framework. This page reads the chapters the act added to the Business and Commerce Code, and the records its defense calls for.
On this page
A reading of the enrolled text, dated and sourced.
In this document, "we" means Hikari Blue, the firm that installs the governance controls cited below on client engagements and hands them over. This is not legal advice. Whether the act applies to a given system is a question for counsel.
Section numbers refer to the Business and Commerce Code as added by section 4 of HB 149 (Acts 2025, 89th Legislature, Regular Session, chapter 1174). Summaries follow the enrolled text and the codified chapter; the official text governs.
Where we map the act onto records and controls, we say so: in our reading. Mapping prepared by Hikari Blue; the Texas attorney general has not reviewed it.
Any person doing business in Texas, a short list of prohibited intents, and duties on government.
Section 551.002 applies the act to a person who promotes, advertises or conducts business in Texas, produces a product or service used by Texas residents, or develops or deploys an AI system in Texas. Section 551.001 defines an AI system as a machine-based system that infers from its inputs how to generate outputs, including content, decisions, predictions or recommendations, that can influence physical or virtual environments.
A deployer is a person who deploys an AI system for use in Texas; a developer is a person who develops one offered or provided in Texas (Section 552.001). The act sets no impact assessment, risk management program or record retention duty for a private company.
| Provision | What it says | Applies to |
|---|---|---|
| 552.052 · Manipulation of human behavior | No AI system developed or deployed in a manner that intentionally aims to incite or encourage self-harm, harm to another person or criminal activity. | Any person |
| 552.055 · Constitutional protection | No AI system developed or deployed with the sole intent to infringe, restrict or impair an individual's rights under the US Constitution. | Any person |
| 552.056 · Unlawful discrimination | No AI system developed or deployed with the intent to unlawfully discriminate against a protected class; a disparate impact alone does not show intent. Insurers subject to unfair discrimination statutes are outside the section, and a federally insured financial institution that complies with banking law is considered compliant. | Any person |
| 552.057 · Sexually explicit content | No AI system developed or distributed with the sole intent of producing or distributing material prohibited by sections 43.26 and 21.165 of the Penal Code, nor one that engages in sexual text conversations while impersonating a child younger than 18. | Any person |
| 552.051 · Disclosure to consumers | A governmental agency that makes available an AI system intended to interact with consumers discloses it before or at the time of interaction, clearly, in plain language and without a dark pattern. A provider of a health care service or treatment that uses AI discloses it to the patient no later than the date the service is first provided. | Governmental agencies; health care providers |
| 552.053 · Social scoring | No AI system used to assign a social score that results in detrimental or disproportionate treatment, or in the infringement of a right. | Governmental entities |
| 552.054 · Biometric identification | No AI system developed or deployed to uniquely identify a person from biometric data, or from images gathered from public sources, without consent, where that would infringe a right. | Governmental entities |
Who enforces it, and the program and council the act creates.
Attorney general · 552.101 to 552.103
The attorney general has exclusive authority to enforce Chapter 552, and the act creates no private right of action. The attorney general keeps an online complaint mechanism, due by 1 September 2026 under section 8 of the act, and may issue a civil investigative demand after a complaint.
Licensing agencies · 552.106
Once a person has been found in violation under Section 552.105 and the attorney general recommends it, the agency that licenses the person may suspend or revoke the license and impose a penalty of up to $100,000.
Regulatory sandbox · Chapter 553
The Department of Information Resources runs a sandbox in which an approved participant tests an AI system for up to 36 months without a state license, reporting quarterly. The prohibitions of Subchapter B, Chapter 552, cannot be waived.
Texas Artificial Intelligence Council · Chapter 554
7 members appointed by the governor, the lieutenant governor and the speaker of the house. The council studies, reports and recommends; it may not adopt binding rules or guidance.
Chapter 552 preempts any local ordinance or rule on the use of AI systems (Section 552.003), and Section 551.001 limits a consumer to a Texas resident acting in an individual or household context.
Section 552.105(e)(2)(D) protects a violation found by internal review, under a recognized framework.
Section 552.105(e) lists the cases in which a defendant may not be found liable. The last one names a framework: the defendant discovers the violation through an internal review process while it substantially complies with the most recent NIST Generative AI Profile, or another nationally or internationally recognized AI risk management framework.
| Subsection | The defendant is not liable if |
|---|---|
| 552.105(e)(1) | another person uses the AI system affiliated with the defendant in a manner the chapter prohibits. |
| 552.105(e)(2)(A) | it discovers the violation through feedback from a developer, deployer or other person who believes a violation has occurred. |
| 552.105(e)(2)(B) | it discovers the violation through testing, including adversarial or red-team testing. |
| 552.105(e)(2)(C) | it discovers the violation by following guidelines set by applicable state agencies. |
| 552.105(e)(2)(D) | it discovers the violation through an internal review process, if it substantially complies with the most recent NIST Generative AI Profile or another nationally or internationally recognized AI risk management framework. |
NIST AI 600-1 · Generative AI Profile
Published by NIST on 26 July 2024: 12 risks unique to or worsened by generative AI, with suggested actions keyed to AI RMF subcategories. The statute names its most recent version, not a dated one.
The other protections of Section 552.105
A rebuttable presumption that the person used reasonable care (c); an expedited hearing or declaratory judgment for a defendant acting in good faith (d); no civil penalty action for an AI system that has not been deployed (f).
The act does not define substantial compliance. In our reading, a defendant shows it with records of the framework in operation at the date of the review, not with a policy adopted after the notice.
A written notice, 60 days to cure, then 3 tiers of civil penalty.
Under Section 552.104, the attorney general notifies the person in writing and names the provisions alleged to be violated. No action may be brought before the 60th day, nor at all if the person cures within that period and sends a written statement with supporting documentation and the internal policy changes made to prevent a further violation.
| Violation not cured | Civil penalty |
|---|---|
| Curable, or a breach of the cure statement · 552.105(a)(1) | $10,000 to $12,000 per violation |
| Uncurable · 552.105(a)(2) | $80,000 to $200,000 per violation |
| Continued · 552.105(a)(3) | $2,000 to $40,000 for each day the violation continues |
The court decides whether a violation is curable. The attorney general may also seek an injunction and recover attorney's fees, court costs and investigative expenses (552.105(b)).
A licensing agency may add its own sanctions, up to $100,000, once a violation is found and on the attorney general's recommendation (552.106).
The defense is shown with records written before the notice arrives.
In our reading, 3 moments call for evidence: the internal review that found the violation, the substantial compliance with a framework on that date, and the cure statement within 60 days. A civil investigative demand can ask for the same facts at any point after a complaint.
Each line below names the field of the 9-field record of an AI audit trail that answers it, and the control of our governance framework that produces it. Mapping prepared by Hikari Blue; the Texas attorney general has not reviewed it.
| What the act asks | Field of the 9-field record | Governance control |
|---|---|---|
| An internal review process found the violation · 552.105(e)(2)(D) | Action, Human verification, Timestamp | GOV-E-03 · GOV-E-04 |
| Substantial compliance with a recognized framework on that date · 552.105(e)(2)(D) | Policy applied | GOV-P-01 · GOV-P-02 |
| Another person used the system in a prohibited manner · 552.105(e)(1) | Actor identity | GOV-A-01 · GOV-A-02 |
| Cure statement with supporting documentation and policy changes · 552.104(b)(2) | Policy applied, Hash chain | GOV-P-02 · GOV-E-01 |
| Purpose, training data, inputs, outputs, metrics, limitations and post-deployment monitoring, on demand · 552.103(b) | Model and version, Inputs reference, Action | GOV-P-01 · GOV-P-04 · GOV-E-01 |
| Whether the system was deployed · 552.105(f) | Model and version, Timestamp | GOV-P-03 |
Controls. GOV-P-01 use-case register and classification; GOV-P-02 policy as code; GOV-P-03 model selection and change control; GOV-P-04 instructions for use and transparency; GOV-A-01 one identity per agent; GOV-A-02 declared scope and budget; GOV-E-01 append-only, chained trail; GOV-E-03 monitoring and incident record; GOV-E-04 quarterly board pack.
The Generative AI Profile keys its suggested actions to AI RMF subcategories, so our NIST AI RMF crosswalk shows which of them the controls evidence directly and which belong to another program.
2 state laws, 2 designs: prohibited intents in Texas, decision records in Colorado.
Texas names a framework and asks for no records. Colorado asks for records and, from 1 January 2027, names no framework.
| Point | Texas HB 149 | Colorado SB26-189 |
|---|---|---|
| In force | Since 1 January 2026 | From 1 January 2027, for consequential decisions made from that date |
| What is regulated | Prohibited intents for any person; listed uses by governmental entities | Automated decision-making technology that materially influences a consequential decision |
| Records required | None; documentation on civil investigative demand (552.103) | At least 3 years after each consequential decision (6-1-1703) |
| Framework named | NIST Generative AI Profile or another recognized framework, in a defense (552.105(e)(2)(D)) | None in the enacted text; SB24-205 named the NIST AI RMF and ISO/IEC 42001 |
| Enforcement and cure | Attorney general alone; 60 days to cure, with a written statement | Attorney general alone; 60 days to cure where a cure is possible, until 1 January 2030 |
| Civil penalties | $10,000 to $200,000 per violation; $2,000 to $40,000 per day of continued violation | Up to $20,000 per violation under the Colorado Consumer Protection Act, according to the final fiscal note |
| Private right of action | None (552.101(b)) | None (6-1-1706(4)) |
In our reading, a company with customers in both states keeps one record per action: Colorado reads it as 3 years of decision records, Texas reads it as the trace of the internal review and of the framework in force.
The questions buyers ask about Texas HB 149.
Is Texas HB 149 in effect?
Yes. HB 149 was signed by the governor on 22 June 2025 and took effect on 1 January 2026, under section 10 of the act. It added Chapters 551 to 554 to the Business and Commerce Code.
Who enforces TRAIGA?
The Texas attorney general, with exclusive authority under Section 552.101; the act creates no private right of action. A licensing agency may add sanctions once a violation is found, on the attorney general's recommendation.
What are the penalties under Texas HB 149?
After a written notice and 60 days to cure, a court may impose $10,000 to $12,000 per curable violation, $80,000 to $200,000 per uncurable violation, and $2,000 to $40,000 for each day a violation continues (Section 552.105(a)).
What is the NIST defense in TRAIGA?
Under Section 552.105(e)(2)(D), a defendant may not be found liable when it discovers a violation through an internal review process while substantially complying with the most recent NIST Generative AI Profile or another nationally or internationally recognized AI risk management framework.
Does TRAIGA require an impact assessment or records?
No. The act sets no impact assessment, risk program or retention duty for a private company. The attorney general may still request documentation through a civil investigative demand, and both the framework defense and the cure statement rest on documentation.
Does TRAIGA apply to a company based outside Texas?
It applies to a person who promotes, advertises or conducts business in Texas, produces a product or service used by Texas residents, or develops or deploys an AI system in Texas (Section 551.002). Whether a given company falls within it is a question for counsel.
Reference documents and primary sources.
Every Texas date and provision on this page comes from the primary sources below, consulted on 25 October 2026. Colorado facts come from our Colorado guide and its sources.
Reference documents
- NIST AI RMF implementation, control by control v1.0 Crosswalk of 21 October 2026; section 4 dates the Texas and Colorado instruments.
- Enterprise AI Governance Framework v1.0 15 controls with owners, cadences, evidence and EU AI Act hooks. CC BY 4.0.
- What is an AI operating layer? The definition, the 4 tests and the regulatory doctrine this guide applies.
- The US does not have an AI Act. It has regulators. Briefing of 16 August 2026 on the US regulatory mosaic: NIST, Colorado, sector regulators, the FTC.
Primary sources, consulted on 25 October 2026
- Texas Legislature Online · HB 149, enrolled text (opens in a new tab) Sections 1 to 10: Chapters 551 to 554 of the Business and Commerce Code, the complaint mechanism and the effective date.
- Texas Legislature Online · HB 149, bill history (opens in a new tab) Signed by the governor on 22 June 2025, effective 1 January 2026.
- Texas Constitution and Statutes · Business and Commerce Code, Chapter 552 (opens in a new tab) The chapter as codified: added by Acts 2025, 89th Legislature, chapter 1174, effective 1 January 2026; Section 552.105(e)(2)(D).
- NIST AI 600-1 · Generative Artificial Intelligence Profile (opens in a new tab) NIST, 26 July 2024. The profile named in Section 552.105(e)(2)(D). PDF.
How to cite this document.
This URL is permanent and always serves the current version. Each change of substance increments the version and is listed in the changelog below, so a citation stays traceable to the text it quoted.
Preferred citation
Hikari Blue, "Texas HB 149, the Texas Responsible AI Governance Act v1.0", 25 October 2026, https://hikariblue.com/united-states/texas-traiga
Section anchors are stable: #status, #scope, #framework-defense, #penalties, #records, #texas-colorado, #faq, #references, #cite.
Changelog
v1.0 · 25 October 2026. First public version.
License. Published under CC BY 4.0: free to cite, quote and reuse with attribution to Hikari Blue and a link to this page. The license covers Hikari Blue’s text, tables and diagrams on this page; it does not cover third-party data, quotations or legal texts, which keep their own terms, nor the Hikari Blue name and logo, trademarks of UNITED4 LLC.
Show the internal review before the attorney general asks for it.
30 minutes with a named partner. We take one AI workflow you run with Texas users, read it against the prohibitions of Chapter 552, and show which records would support the defense of Section 552.105(e) today. You leave with the list, whether or not we ever work together.
- One workflow read against the prohibitions of Chapter 552
- The records that support the defense today, and the ones missing
- Who should own each gap