Texas AI Law

Guide

Who Is Responsible When AI Gets It Wrong

When an AI output causes harm, ordinary Texas law decides who pays.

Law checked through

Short Answer

AI-related harm can produce contract, consumer, negligence and other claims with their own elements. The Texas Responsible Artificial Intelligence Governance Act (TRAIGA) supplies targeted public enforcement and no private cause of action. Business and Commerce Code § 552.101(b) also bars using its violation as the basis for a private action under another law; an independent claim must stand on that law’s elements. Texas proportionate-responsibility rules can allocate responsibility among covered parties when the chapter applies. Whether a software use fits a products-liability theory and whether Section 230 applies depend on the product, content and allegations. A vendor contract does not bind an injured nonparty to the buyer’s chosen allocation.

Which Laws Apply

Texas AI-specific: Business and Commerce Code § 552.052 and Business and Commerce Code § 552.101 (no private action); Business and Commerce Code § 552.105(e)(1) (misuse defense in Attorney General actions).

Generally applicable Texas law: negligence and negligent misrepresentation; Deceptive Trade Practices-Consumer Protection Act (DTPA), Business and Commerce Code §§ 17.41 to 17.63; products liability, Civil Practice and Remedies Code chapter 82; proportionate responsibility, Civil Practice and Remedies Code chapter 33; Uniform Commercial Code warranties for goods.

Federal: 47 U.S.C. § 230; Federal Trade Commission (FTC) Act.

Theories of Liability

TheoryRequired inquiry
NegligenceRecognized duty, breach, causation and damages. A wrong output alone does not prove every element.
Negligent misrepresentationFalse information supplied in the course of business or a transaction with a pecuniary interest, lack of reasonable care, justifiable reliance and pecuniary loss (Federal Land Bank Ass’n of Tyler v. Sloane, 825 S.W.2d 439, 442 (Tex. 1991), adopting Restatement (Second) of Torts § 552). The same decision limits recovery to pecuniary loss, excluding benefit-of-the-bargain and mental anguish damages, and treats the claim as sounding in tort rather than contract.
DTPAConsumer standing and an actionable provision of § 17.50; listed deceptive-act claims require reliance. Enhanced relief depends on the finding: a knowing violation supports mental anguish damages plus up to three times economic damages, while an intentional violation supports mental anguish damages plus up to three times the combined mental anguish and economic damages (Business and Commerce Code § 17.50(b)(1)).
Products liabilityCovered product and defect theory, causation and the provisions of chapter 82. Pure software and generated information require classification analysis.
Contract or warrantyActual promise, applicable contract, performance, causation and valid limitations. UCC goods coverage must be established rather than assumed for all software.

Is AI a Product?

Civil Practice and Remedies Code chapter 82 regulates products-liability actions through its definitions and remedies. It does not expressly decide every form of generative software or chatbot output. A products claim must identify the product, manufacturer or seller, alleged defect and harm. Contract and consumer claims need separate analysis even where product classification is disputed.

Allocating Fault

Chapter 33 lets a jury assign percentages of responsibility to each defendant, settling party and designated responsible third party. In an AI case, that may include the developer, the business that deployed the system, an integrator and the user. Contracts can shift some of that exposure between businesses, subject to Texas’s express negligence and fair notice rules for indemnities.

Section 230

Section 230(c)(1) concerns treatment of a provider or user as publisher or speaker of information provided by another information-content provider. Whether a claim concerns third-party content, the provider’s own generated content or an independent design or business duty matters. It is not a blanket immunity for everything an AI assistant says or does. Do not assume a generated response meets the third-party-content condition without analyzing the system and claim.

Regulatory Enforcement

The Texas Attorney General’s September 2024 Pieces assurance applies the DTPA to questioned AI accuracy marketing and records negotiated commitments, not adjudicated damages. TRAIGA’s misuse defense in Business and Commerce Code § 552.105(e)(1) belongs to the specified Attorney General action. It does not automatically resolve an independently supported contract, privacy or consumer claim.

Illustrative Example (Hypothetical)

Hypothetical: a Texas property manager’s assistant misstates a repair obligation and a tenant incurs a loss. A contract or consumer claim depends on the lease, transaction, standing, reliance and harm. The vendor agreement may allocate defense or reimbursement obligations between the businesses; that is distinct from contribution under the governing liability rules. TRAIGA supplies no private action for the misstatement.

What Is Unsettled

Whether AI software is a product under chapter 82; whether Section 230 protects generated output; what standard of care applies to deploying AI in different settings.

Sources