Texas Responsible Artificial Intelligence Governance Act (HB 149)

What it is

The Texas Responsible Artificial Intelligence Governance Act (TRAIGA) is House Bill 149 of the 89th Texas Legislature. It was signed on 22 June 2025 and took effect on 1 January 2026. It adds Subtitle D, Artificial Intelligence Protection (Chapters 551 to 554), to Title 11 of the Texas Business and Commerce Code, and amends the state’s biometric identifier and consumer privacy laws.

The enacted Act is much narrower than the first TRAIGA bill, HB 1709, which followed the Colorado model of duties for high-risk systems. What passed is a set of prohibitions defined by the developer’s or deployer’s intent, disclosure duties for government and health care, an Attorney General enforcement regime with a cure period, a regulatory sandbox and a Texas Artificial Intelligence Council.

Who it applies to

The Act 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. Out-of-state companies whose AI reaches Texans are inside it. Most prohibitions apply to any developer or deployer. A separate set applies only to governmental entities.

What it requires

Prohibitions for everyone. A person may not develop or deploy an AI system in any of the following ways.

• in a way that intentionally aims to incite or encourage a person to harm themselves, harm another person or engage in criminal activity
• with the sole intent of infringing, restricting or impairing a person’s rights under the US Constitution
• with the intent to unlawfully discriminate against a protected class in violation of state or federal law (disparate impact alone does not show intent)
• with the sole intent of producing or distributing child sexual abuse material or unlawful sexual deepfakes, or to run text conversations that simulate sexual conduct while impersonating a minor

Insurers subject to insurance anti-discrimination law, and federally insured financial institutions that comply with banking law, are treated separately under the discrimination prohibition.

Prohibitions for government. Governmental entities may not use AI for social scoring that leads to unjustified or unrelated detrimental treatment, and may not use AI to uniquely identify a person from biometric data or images scraped from public sources without consent where that would infringe the person’s rights.

Disclosure. A governmental agency that makes an AI system available to interact with consumers discloses, before or at the time of interaction, that the consumer is dealing with AI, even where that would be obvious. The disclosure must be clear, conspicuous and in plain language, without dark patterns. Where AI is used in health care services or treatment, the provider discloses it to the patient or their representative by the date the service is first provided, or as soon as reasonably possible in an emergency.

Enforcement. The Attorney General has exclusive enforcement authority and runs an online complaint mechanism. On a complaint it can issue a civil investigative demand asking for a description of the system’s purpose and intended use, the training data, inputs and outputs, performance metrics, known limitations, post-deployment monitoring and safeguards. Before suing, it must give written notice of the provisions it says were breached. The person then has 60 days to cure and give a written statement, with supporting documentation, of the cure and of the internal policy changes made to prevent a repeat. There is a rebuttable presumption that a person used reasonable care. No action can be brought for a system that has not been deployed.

Defences. A defendant is not liable where another person misuses its system in a prohibited way, or where it discovers a violation through feedback from others, adversarial or red-team testing, following state agency guidelines, or an internal review process while substantially complying with the NIST AI RMF Generative AI Profile or another recognised AI risk management framework.

Sandbox and council. The Department of Information Resources runs a regulatory sandbox in which approved participants can test AI systems for up to 36 months without certain state licences, with quarterly reporting. The Texas Artificial Intelligence Council advises the Legislature and agencies.

Key dates

• 22 June 2025. HB 149 signed.
• 1 January 2026. The Act takes effect.

How it relates to other frameworks

TRAIGA is the clearest case in US law of a statute rewarding a voluntary framework. The defence names the NIST AI RMF Generative AI Profile, which builds on the NIST AI RMF, and accepts other nationally or internationally recognised frameworks, which on the plain wording could include ISO/IEC 42001. It does not excuse a violation in advance. It protects an organisation whose own testing or review finds the problem, which only works if that testing and review actually happen and leave a record.

The intent-based prohibitions overlap with the EU AI Act’s banned practices on manipulation and social scoring, though the EU bans those practices by effect rather than intent. The government disclosure duty resembles Utah’s regulated-occupation disclosure. Texas’s consumer privacy law, the Texas Data Privacy and Security Act, separately gives residents an opt-out of profiling, covered on the US state privacy laws on profiling page.

Executive Order 14365 of December 2025 directed the US Attorney General to challenge state AI laws considered burdensome. It does not itself pre-empt TRAIGA, and Congress has not passed pre-emption.

Working with it in Aicura

US state framework content is available in Aicura on the Enterprise tier and is switched on in Settings under Frameworks. Risk assessments for each register entry produce risk items with written rationale and suggested controls, which land in Risks and Controls to be tracked as implemented and tested. Security testing records adversarial and red-team runs, deliberate attempts to make your systems and agents misbehave, and guardrails enforce rules on what models may do at the AI gateway that sits between your applications and the models they call. Together they give you the testing and internal review the Act’s defence describes, and the Evidence Vault, where Aicura seals timestamped snapshots so anyone can confirm they have not been changed, holds a snapshot of when each happened. Aicura gives guidance for your own judgement and your advisers’. It does not decide whether a use is prohibited.