Texas Responsible AI Governance Act (TRAIGA / HB 149)

The Texas Responsible Artificial Intelligence Governance Act, enacted as House Bill 149 of the 89th Legislature, Regular Session, is Texas's comprehensive AI governance statute. It pairs a set of intent-based prohibited-use rules with a statutory regulatory sandbox and a non-binding advisory council, and is enforced exclusively by the state attorney general subject to a 60-day cure period. It was the first comprehensive AI law enacted in a Republican-led state and the second state-level comprehensive AI statute, after Colorado.

Status and timeline

TRAIGA was filed in the 89th Legislature, Regular Session (89R) in 2025, signed into law by Governor Greg Abbott on June 22, 2025, and took effect January 1, 2026. The enacting body was the Texas Legislature.

Scope and definitions

Under §551.002, TRAIGA applies to any 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. This is a broad commerce-clause-style trigger, similar to California's CCPA and broader than the developer-only triggers in California SB 53 — Transparency in Frontier AI Act and the New York RAISE Act (S. 8828).

Key provisions

Prohibited practices (Subchapter B, §§552.051–552.057)

The core of the statute is a set of intent-based prohibitions, in contrast to Colorado's duty-of-care model and the EU's risk-tier classification.

Under §552.052, AI systems that "intentionally aim to incite or encourage" physical self-harm, harm to others, or criminal activity are prohibited. Under §552.053, Texas governmental entities may not use AI to evaluate persons based on "social behavior" in ways that result in "detrimental or unfavorable treatment" or infringement of constitutional or legal rights; this tracks the EU AI Act Article 5(1)(c) prohibition but limits it to government uses. Under §552.055, AI systems developed or deployed "with the sole intent ... to infringe, restrict, or otherwise impair" US constitutional rights are prohibited; the sole-intent standard is narrow and likely difficult to enforce.

Under §552.056, AI used "with the intent to unlawfully discriminate against a protected class" is prohibited. The statute states that "disparate impact is not sufficient by itself" to prove intent, a departure from Colorado, which treats disparate-impact evidence as central to algorithmic-discrimination claims. Under §552.057, AI systems developed or distributed with intent to produce visual material violating Penal Code §43.26, or to engage in sexualized text conversations "while impersonating or imitating a child younger than 18 years of age," are prohibited.

Under §552.051, state agencies must disclose AI use before or at the time of interaction; healthcare interactions must be disclosed no later than first service.

Regulatory sandbox (Chapter 553)

TRAIGA establishes the first statutory US AI sandbox. It allows up to 36 months of testing (§553.053), extendable for "good cause." During testing, the attorney general may not file or pursue charges for violations of waived laws (§553.051(c)). The Subchapter B prohibitions (manipulation, discrimination, CSAM) cannot be waived (§553.051(e)). The sandbox has no analog in Colorado, California, New York, or the EU AI Act.

Texas AI Advisory Council (Chapter 554)

The council has seven members (§554.002(a)): three appointed by the Governor, two by the Lieutenant Governor, and two by the Speaker. Members must be Texas residents with AI, data privacy, ethics, public policy, risk management, governmental efficiency, or antitrust expertise (§554.003). Its authority is limited to reports and training; it "may not adopt rules or promulgate guidance that is binding" (§554.103(1)). This non-binding authority contrasts with Colorado's grant of rulemaking power to the state attorney general.

Enforcement and penalties (Subchapter C)

Enforcement is exclusive to the attorney general, with no private right of action (§552.101). A 60-day cure period applies (§552.104): after written notice, the attorney general cannot sue for 60 days if the violator cures the violation and submits written certification with supporting documentation and policy changes; this mirrors Colorado's cure mechanism. Civil penalties (§552.105) are set at $10,000–$12,000 per violation for curable violations, $80,000–$200,000 per violation for uncurable violations, and $2,000–$40,000 per day for continuing violations. Injunctive relief, attorney fees, and court costs are also available.

Comparison with other AI laws

Relationship to the Colorado template

TRAIGA was originally modeled on Colorado SB 24-205 but emerged substantially scaled back. It requires intentional prohibited conduct, whereas Colorado requires reasonable care to avoid algorithmic discrimination regardless of intent. It imposes no duty of care: Colorado requires impact assessments, risk management, and consumer notice, while TRAIGA imposes no affirmative obligations outside the prohibited-use list. TRAIGA explicitly states that disparate impact alone cannot prove discrimination, while Colorado treats it as central. Colorado's attorney general can issue implementing rules; Texas's council is advisory only. Texas adds a 36-month testing program that Colorado lacks. Because the prohibitions are intent-gated, the penalty schedule has a smaller surface area than Colorado's. The result is a narrower analog of the Colorado frame that mirrors the EU's Article 5 prohibitions while not adopting the EU's high-risk duty-of-care apparatus.

Reception and context

TRAIGA was the first comprehensive AI law enacted in a Republican-led state, indicating that comprehensive AI statutes were not confined to Democratic-led states. Commentators have noted that its intent-based design may be less exposed to First Amendment challenges than effects-based regimes: xAI's lawsuit against Colorado (see AI and the First Amendment) argues compelled speech, and TRAIGA's intent standard and explicit disparate-impact disclaimer may be less vulnerable to that line of attack. It is the first US AI statute with a formal regulatory sandbox, a feature other states may adopt. The statute is an instance of Techno-Federalism: How Regulatory Fragmentation Shapes the U.S.-China AI Race, with Texas legislating in the absence of a comprehensive federal AI framework under a distinct political framing from Colorado, California, and New York.

Relationships

• depends-on: Colorado AI Act (SB 24-205) and SB 25B-004 (Date Amendment) — original template
• contradicts: Colorado AI Act (SB 24-205) and SB 25B-004 (Date Amendment) — on disparate-impact evidentiary standard
• related: California SB 53 — Transparency in Frontier AI Act, New York RAISE Act (S. 8828) — contrasting frontier-transparency approach
• related: EU AI Act (Regulation 2024/1689) — structural parallel to Article 5 prohibitions
• related: AI and the First Amendment — intent-based design and constitutional-challenge posture
• related: Techno-Federalism: How Regulatory Fragmentation Shapes the U.S.-China AI Race
• related: US AI Regulatory Approaches Compared — adds a ninth regulatory approach

Sources

• Texas TRAIGA source summary
• Raw text: Texas Responsible AI Governance Act (TRAIGA / HB 149)
• Texas Legislature Online: https://capitol.texas.gov/tlodocs/89R/billtext/html/HB00149F.htm