Texas wrote its own AI statute after investigating chatbots

The design

Texas's legislature passed House Bill 149, the “Texas Responsible Artificial Intelligence Governance Act” (TRAIGA), and the governor signed it on 22 June 2025. The bill's own history page at the Texas Legislature records that signing date and shows the act's own effective date of 1 January 2026. Unlike the Texas Attorney General's 2024 investigation of Character.AI and Meta AI, which used existing consumer-protection authority to examine specific companies, TRAIGA is new statutory text: it adds Chapters 551 through 553 to the Business and Commerce Code, defining “developer,” “deployer,” and duties around artificial intelligence systems generally, not any single company's product.

What the evidence says

The enrolled bill text is the primary record of what the act requires, and it is narrower than a comprehensive AI statute. Section 552.052, titled “Manipulation of Human Behavior,” states that a person may not develop or deploy an AI system “in a manner that intentionally aims to incite or encourage a person to... commit physical self-harm, including suicide... harm another person; or... engage in criminal activity.” A separate section requires disclosure, clear and in plain language, when a consumer interacts with an AI system in specified contexts such as health care, and bars using a dark pattern to obscure that disclosure. The act also preempts local ordinances that would otherwise regulate AI use.

What it asks of people

TRAIGA asks a developer or deployer to build in a disclosure and to refrain from designing a system toward the listed harms; it does not give an individual consumer a private right to sue over a violation. Enforcement runs through the Texas Attorney General, who may seek civil penalties. For a user of a companion-style chatbot operating in Texas, the practical effect is that the disclosure duty and the manipulation prohibition apply regardless of whether the product is marketed as a companion, a coach, or a general assistant, so long as it meets the act's definitions.

Privacy and safeguards

The enrolled text does not set data retention or deletion rules; it leaves privacy obligations to Texas's separate data privacy statute and addresses AI specifically as a behavior and disclosure regime. Its self-harm and manipulation provision is the clearest safeguard relevant to companion apps, but it is framed around intentional design rather than around detecting distress in a single conversation, and the act does not itself require crisis-resource referrals the way some companies' policies do voluntarily. The bill's construction clause also states the chapter may not be read to burden free speech rights, a limit courts have not yet interpreted.

• Does the disclosure duty in Chapter 552 apply to a specific companion product, given its stated exceptions?
• How will “intentionally aims to incite” in the manipulation section be proven in an actual enforcement action?
• What guidance has the Texas Attorney General published since 1 January 2026 interpreting these duties?

TRAIGA is a statute built independently of the Attorney General's earlier company-specific inquiries, and its own text, not that earlier investigation, is what governs a Texas-based companion product from 2026 onward.