Texas AI Law in 2026: What HB 149 (TRAIGA) Prohibits, Who It Reaches, and How the Attorney General Enforces It

What does the Texas AI law require?

Texas HB 149 prohibits developing or deploying an AI system that intentionally aims to incite or encourage self-harm, harm to another person, or criminal activity, or with the intent to unlawfully discriminate against a protected class, or with the sole intent to infringe constitutional rights. It also prohibits developing or distributing an AI system with the sole intent of producing certain sexual material involving minors or unlawful deepfakes, or one that intentionally simulates sexual conduct while imitating a child. It requires governmental agencies to tell consumers when they are interacting with an AI system, and it requires health care providers to disclose AI used in a service or treatment by the date the service is first provided, or as soon as reasonably possible in an emergency. The Attorney General enforces it after written notice and a 60-day cure period, with civil penalties from $10,000 to $200,000 per uncured violation and $2,000 to $40,000 for each day a violation continues.

It requires no impact assessment, risk management program, bias audit, or general disclosure from private companies, and Chapter 552 supersedes and preempts any ordinance, resolution, rule, or other regulation a political subdivision adopts on the use of AI systems (Sec. 552.003). The governor signed HB 149 on June 22, 2025 (Acts 2025, 89th Leg., R.S., Ch. 1174).

Who is covered by the Texas AI law?

Chapter 552 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 (Sec. 551.002). A developer is a person who develops an AI system that is offered, sold, leased, given, or otherwise provided in Texas; a deployer is a person who deploys an AI system for use in Texas (Sec. 552.001).

The prohibitions on manipulation, constitutional infringement, unlawful discrimination, and sexual content bind any person. The social scoring limit and the Sec. 552.054(b) biometric identification limit bind governmental entities, and the general consumer disclosure binds governmental agencies, a term Chapters 551 and 552 do not define. Sec. 552.054(c) also makes any violation of the state biometric identifier statute (Sec. 503.001), which binds any person capturing a biometric identifier for a commercial purpose, a violation of Sec. 552.054. A “governmental entity” is a department, commission, board, office, authority, or other administrative unit of the state or a political subdivision that exercises governmental functions, excluding hospital districts and institutions of higher education (Sec. 552.001(3)). A “consumer” is a Texas resident acting in an individual or household context, not a commercial or employment one (Sec. 551.001(2)).

What does HB 149 prohibit?

Each of the four prohibitions that bind any person turns on intent, so the statute is narrower than it reads. For a private company, the discrimination bar shows how much narrower.

The discrimination bar is on developing or deploying a system “with the intent to unlawfully discriminate against a protected class,” and “a disparate impact is not sufficient by itself to demonstrate an intent to discriminate” (Sec. 552.056(b) and (c)). A hiring or lending model with a disparate outcome does not violate HB 149 on that fact alone. It can still create exposure under Title VII’s disparate impact provision, the Fair Housing Act, state laws that recognize disparate impact, or California’s employment regulations. The CFPB’s Regulation B rule, in effect since July 21, 2026 and under court challenge, states that ECOA does not authorize disparate impact liability. Colorado SB 26-189 imposes procedural duties, not a disparate impact standard.

What must government agencies and health care providers disclose?

A governmental agency that makes available an AI system intended to interact with consumers must disclose to each consumer, before or at the time of interaction, that the consumer is interacting with an AI system (Sec. 552.051(b)). The disclosure is required even where it would be obvious to a reasonable consumer. It must be clear and conspicuous, in plain language, without a dark pattern, and it may be provided through a hyperlink to a separate web page (Sec. 552.051(c) to (e)).

Subsection (f) extends that disclosure to health care. If an AI system is used in relation to health care service or treatment, the provider must give the disclosure to the recipient or the recipient’s personal representative not later than the date the service or treatment is first provided, except in an emergency, when the provider must give it as soon as reasonably possible. The subsection names the provider, not the governmental agency, so private hospitals and practices are covered.

How do enforcement and the cure period work?

The Attorney General has exclusive authority to enforce Chapter 552, apart from licensing sanctions under Sec. 552.106, and the chapter provides no private right of action (Sec. 552.101). Section 8 of the Act required an online complaint mechanism by September 1, 2026; the Attorney General’s Consumer AI Rights page carried a “File an AI Complaint” route as of October 23, 2026. A complaint received through that mechanism is what lets the Attorney General issue a civil investigative demand for documentation on the system, from training data type to post-deployment monitoring (Sec. 552.103).

Before suing, the Attorney General must send written notice identifying the provisions allegedly violated. No action may be brought before the 60th day after notice, or at all if, before that day, the person cures the violation and delivers a written statement that it has cured it, with supporting documentation and any necessary changes to internal policies to reasonably prevent recurrence (Sec. 552.104).

For an uncured violation the civil penalty is $10,000 to $12,000 per violation the court finds curable or per breach of a cure statement, $80,000 to $200,000 per violation the court finds uncurable, and $2,000 to $40,000 for each day a violation continues, plus an injunction and fees (Sec. 552.105(a) and (b)). There is a rebuttable presumption of reasonable care, and no penalty action may be brought for a system that has not been deployed (Sec. 552.105(c) and (f)). If a court has found a licensee in violation under Sec. 552.105 and the Attorney General recommends additional enforcement, the licensing agency may add suspension, probation, or revocation and a penalty of up to $100,000 for a violation of Subchapter B (Sec. 552.106).

How does the framework safe harbor work?

Under Sec. 552.105(e), a defendant may not be found liable if another person used the defendant’s AI system in a prohibited manner, or if the defendant discovered the violation through feedback from a developer, deployer, or other person; through testing, including adversarial or red-team testing; by following guidelines set by applicable state agencies; or through an internal review process, if the defendant substantially complies with the most recent version of the NIST Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile or another nationally or internationally recognized AI risk management framework.

The framework clause attaches to the internal review route. A company that adopts the NIST profile but never runs the internal review that finds the problem gets nothing from the framework alone. A company that finds the problem through red-team testing has the defense with or without a framework.

What are the sandbox program and the Texas Artificial Intelligence Council?

Chapter 553 directs the Department of Information Resources, in consultation with the council, to run a regulatory sandbox in which a person may test an innovative AI system without a license or other regulatory authorization (Sec. 553.051). An applicant needs approval from the department and any applicable agency, on a department form that requires a detailed description and intended use of the system, a benefit assessment, a plan for mitigating adverse consequences, and proof of compliance with applicable federal AI laws (Sec. 553.052). Participation runs up to 36 months, extendable for good cause, with quarterly participant reports (Secs. 553.053 and 553.102). The Attorney General and state agencies may not pursue a participant for violating a waived law during the testing period, but the duties and prohibitions in Chapter 552, Subchapter B cannot be waived and remain enforceable (Sec. 553.051(c) to (e)).

Chapter 554 creates the Texas Artificial Intelligence Council: seven public members, three appointed by the governor, two by the lieutenant governor, two by the speaker, on staggered four-year terms (Sec. 554.002). It may issue reports to the legislature, must train state agencies and local governments, and may not adopt binding rules (Secs. 554.001 to 554.103).

The department’s page on 89th Legislature AI laws, dated August 14, 2026, says only that HB 149 created the council and a regulatory sandbox; it lists no appointments, meetings, or application window. The department’s published sandbox page covers a different program, the Public Sector AI Sandbox under SB 1964. This council is not the Public Sector Artificial Intelligence Systems Advisory Board appointed in November 2025.

How does HB 149 compare with Colorado SB 26-189 and California?

Texas regulates intent. Colorado’s law and California’s privacy rules regulate the process around consequential or significant decisions.

The state AI laws hub carries the Colorado detail and the preemption fight; the California spoke covers each California statute.

What other Texas AI laws are in force?

Two health care statutes from the same session and a biometric amendment inside HB 149 also govern AI use in Texas.

• SB 1188 (Health and Safety Code Chapter 183), effective September 1, 2025. A practitioner may use AI for diagnostic purposes, including recommendations on a diagnosis or course of treatment, if acting within the scope of the practitioner’s license, certification, or other authorization, if the use is not otherwise restricted or prohibited by state or federal law, and if the practitioner reviews all records created with AI in a manner consistent with medical records standards developed by the Texas Medical Board. The practitioner must also disclose that use to patients (Sec. 183.005).
• SB 815, effective September 1, 2025. A utilization review agent may not use an automated decision system to make, wholly or partly, an adverse determination under a health benefit plan delivered, issued, or renewed on or after January 1, 2026; administrative support and fraud-detection uses are not prohibited. See the insurers page.
• HB 149 Section 2, amending the biometric identifier statute (Sec. 503.001). Consent to capture a biometric identifier for a commercial purpose cannot rest solely on an image existing on the Internet unless the individual made it public. The training, processing, or storage of biometric identifiers to develop, train, evaluate, disseminate, or offer AI models or systems is exempt unless a system is used or deployed to uniquely identify a specific individual, and AI used for security, fraud, and illegal-activity purposes is exempt. An identifier captured to train AI and later used for another commercial purpose falls back under the statute’s retention, destruction, and penalty provisions (Sec. 503.001(e) and (f)).

What should a developer or deployer in Texas do now?

• Inventory every AI system offered or used in Texas and tag each as developer, deployer, or both
• Write a one-page intended-use statement per system that rules out the Sec. 552.052 to 552.057 purposes
• Assemble the documentation a civil investigative demand can request under Sec. 552.103(b)
• Place the health care disclosure in intake flows so it reaches the patient or personal representative by the first service date
• Schedule adversarial or red-team testing and an internal review cycle, and log what each finds and when
• Substantially comply with the most recent version of the NIST Generative AI Profile or another nationally or internationally recognized AI risk management framework, and run the internal review under it, so the Sec. 552.105(e) internal review route is available
• Write a 60-day cure playbook: who receives Attorney General notice, who documents the fix, who signs the cure statement

How this fits the other guides: this page is the Texas spoke of the US state AI laws hub. The governance frameworks guide explains the NIST profile the safe harbor names, the hospitals page covers the health care disclosure next to SB 1188, and the employers page covers hiring and HR tools.

This guide is general information as of October 2026, not legal advice; confirm the current text with the Texas Legislature, the Attorney General, or counsel before acting on it.