EdLaw
Texas TRAIGA — Business & Commerce Code Subtitle D (Artificial Intelligence Protection)
H.B. 149 (89th Legislature, 2025), cited by its own §1 as the Texas Responsible Artificial Intelligence Governance Act, in force 1 January 2026. It creates Subtitle D of Title 11 of the Business & Commerce Code: general provisions and definitions (Chapter 551), the substantive duties and prohibitions with their enforcement (Chapter 552), a regulatory sandbox programme (Chapter 553) and the Texas Artificial Intelligence Council (Chapter 554).
TRAIGA is intent-based, not risk-tier-based: unlike the EU AI Act, it does not classify systems by risk and impose graduated conformity duties. It prohibits specified purposes — behavioural manipulation, social scoring, biometric identification without consent, intentional discrimination, and certain sexually explicit content — and leaves everything else to general law. §552.056(c) makes the point explicitly: a disparate impact is not by itself sufficient to demonstrate an intent to discriminate.
Applicability to EdTech
- Subtitle-wide scope is broad: it 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 (§551.002). No revenue or headcount threshold.
- The definition of an "artificial intelligence system" is wide — any machine-based system that, for any explicit or implicit objective, infers from its inputs how to generate outputs including content, decisions, predictions or recommendations that can influence physical or virtual environments (§551.001(1)). A learning product's recommendation or adaptive-sequencing engine is inside this definition.
- Read each duty's own subject carefully — several bind government, not vendors. §552.051 (disclose to the consumer that they are interacting with an AI system) is addressed to "a governmental agency", and §552.054(b) (developing or deploying an AI system to uniquely identify an individual from biometric data or from targeted/untargeted gathering of images from the Internet, without consent) is addressed to "a governmental entity". Texas school districts are governmental entities and political subdivisions — so where a district deploys our AI features, these are the district's duties, and the product has to let the district discharge them (an AI-interaction disclosure surface, and consent handling for any identification feature).
- The duties that bind any person in scope are §552.052 (developing or deploying a system that intentionally aims to incite or encourage self-harm, harm to another, or criminal activity), §552.056 (intentional unlawful discrimination against a protected class) and §552.057 (AI systems for prohibited sexual material, deep fakes, or text-based sexual conversation impersonating a child). §552.053 (social scoring) is a third governmental-entity duty, alongside §552.051 and §552.054(b).
- The CUBI interlock is the practical headline for hardware: §552.054(c) provides that "a violation of Section 503.001 is a violation of this section" (§552.054). A biometric-capture failure under CUBI therefore also becomes a Chapter 552 violation, exposed to this chapter's higher penalties — while the same Act wrote CUBI's new AI carve-outs. The two must be read together, not separately.
- TRAIGA's "biometric data" definition (§552.054(a)) matches the TDPSA one: eye retina or iris in scope; physical or digital photographs, video and audio recordings and data generated from them expressly out.
Contents
| Document | Covers |
|---|---|
| AI-PROTECTION.md | Chapter 551 in full (§§551.001–551.003) and Chapter 552 in full (§§552.001–552.106) |
Chapters 553 (sandbox programme) and 554 (AI Council) are institutional machinery and are linked under Sources rather than extracted: sandbox testing is entered by application and prior approval (§553.052), and the Council "may not … adopt rules or promulgate guidance that is binding for any entity" (§554.103). Extract them if a sandbox application or a Council consultation becomes live work.
Key obligations at a glance
- Governmental agency: disclose AI interaction, clearly, in plain language, without dark patterns — §552.051.
- No AI system developed or deployed to intentionally incite or encourage self-harm, harm to another or criminal activity — §552.052.
- Governmental entity: no social scoring producing the listed detrimental or rights-infringing outcomes — §552.053.
- Governmental entity: no AI system developed or deployed to uniquely identify an individual from biometric data or scraped media without consent — and a CUBI violation is a violation of this section — §552.054.
- No AI system developed or deployed with the intent to unlawfully discriminate against a protected class; disparate impact alone is not intent — §552.056.
- Prohibited sexually explicit content and child pornography — §552.057.
Enforcement
Exclusively the Attorney General (§552.101), with investigative authority under §552.103 and a mandatory 60-day notice and cure period before any action (§552.104) — cure plus a written statement (documentation of the cure and any necessary internal-policy changes) bars the action. There is no private right of action (§552.101(b)), and a rebuttable presumption that the person used reasonable care (§552.105(c)). Civil penalties are tiered by curability: $10,000–$12,000 per curable violation (or per breach of a cure statement), $80,000–$200,000 per uncurable violation, and $2,000–$40,000 per day for a continuing violation (§552.105(a)). Separately, a state agency may sanction its own licensees — suspension, probation or revocation, plus a monetary penalty up to $100,000 — but only after a §552.105 finding and on the Attorney General's recommendation (§552.106). Local regulation of AI is preempted (§552.003).
Licence
The reproduced text is Tex. Bus. & Com. Code chs. 551 and 552, enacted by the Texas Legislature and served by the Texas Legislative Council on statutes.capitol.texas.gov. No copyright subsists in this text. In Georgia v. Public.Resource.Org, Inc., 590 U.S. 255 (2020), the Supreme Court held: "Under what has been dubbed the government edicts doctrine, officials empowered to speak with the force of law cannot be the authors of—and therefore cannot copyright—the works they create in the course of their official duties." It went on: "judges—and, we now confirm, legislators—may not be considered the 'authors' of the works they produce in the course of their official duties as judges and legislators. That rule applies regardless of whether a given material carries the force of law." (slip opinion, pp. 1 and 5–6, read 2026-10-01.)
The reproduced pages carry no copyright notice or terms of use (checked 2026-10-01 against the live pages or the builder's cached copies).
Public-domain text carries no conditions. The corpus attributes it all the same: the {.meta} provenance line on each section names the publisher, the Sources section links the official text, and the Meta section describes what the builder changed (markup and anchors only). Nothing here suggests that the Texas Legislature endorses the corpus.
Related
- Texas CUBI — §552.054(c) folds CUBI violations into this chapter; H.B. 149 also amended CUBI
- Texas TDPSA — §541.104 was amended by the same Act; §552.051 uses its "dark pattern" definition
- EU AI Act — the risk-tiered comparator; TRAIGA is intent-based instead