Texas HB 149 (TRAIGA)

The Texas Responsible Artificial Intelligence Governance Act (TRAIGA) applies broadly to any entity operating in Texas, creating products for Texas residents, or deploying AI systems within the state. This gives the law significant reach across the nation's second-largest economy.

The law defines an 'artificial intelligence system' broadly as any machine-based system that infers from inputs to generate outputs (content, decisions, predictions, recommendations) that can influence physical or virtual environments. This encompasses most modern AI technologies, including generative AI and automated decision tools.

Crucially, 'consumer' is defined as a Texas resident acting 'only in an individual or household context,' explicitly excluding employment and commercial uses from the law's consumer protection provisions. This narrows the focus of certain aspects of the law to personal consumer interactions.

TRAIGA establishes the Texas Artificial Intelligence Advisory Council, a seven-member body appointed by the governor, lieutenant governor, and speaker of the house. This Council serves as the primary governance body for AI in Texas.

The Council's main responsibilities include developing and conducting AI training programs for state agencies and local governments. It is also tasked with issuing reports on various AI-related topics, such as data privacy, security, AI ethics, and legal compliance. These reports are intended to inform the public and government on best practices and emerging issues.

Significantly, the Council is expressly prohibited from promulgating binding rules or regulations. This structural limitation means the Council acts purely as an advisory and educational body, distinguishing Texas's approach from jurisdictions where AI governance bodies have direct regulatory authority.

TRAIGA establishes a 36-month regulatory sandbox, administered by the Department of Information Resources (DIR) in consultation with the AI Advisory Council. This sandbox provides a controlled environment designed to foster AI innovation and experimentation.

The purpose of the sandbox is to allow developers and deployers to test new AI applications with a reduced regulatory burden. This approach aims to encourage technological advancement while still providing a framework for oversight.

DIR will be responsible for developing the specific criteria for participation and the oversight procedures for entities operating within the sandbox. This model is similar to those used in financial services regulation, offering a pathway for novel AI solutions to be developed and tested.

TRAIGA explicitly prohibits the development or deployment of AI systems designed to intentionally encourage any person to physically harm themselves or others. This provision targets AI applications that could be used to manipulate individuals into dangerous actions.

The focus is on the 'intentional' aspect of encouragement, meaning the system's design or purpose must be to provoke such harmful behavior. This is a direct prohibition against AI systems that could be weaponized for psychological manipulation leading to physical harm.

This prohibition underscores a fundamental ethical boundary for AI development and deployment in Texas, prioritizing public safety and individual well-being.

The Act prohibits the deployment of AI systems intended to discriminate against protected classes, including race, sex, and disability. This provision aims to prevent AI from being used as a tool for targeted prejudice.

An important clarification in the law is that disparate impact alone does not establish intent to discriminate. This means that if an AI system's output unintentionally leads to different outcomes for protected groups, it does not automatically constitute a violation, unless discriminatory intent can be proven.

This focus on 'intent' sets a higher bar for proving a violation compared to some other anti-discrimination laws that consider disparate impact sufficient for a finding of discrimination.

TRAIGA enacts strict prohibitions against AI systems involved in child sexual abuse material (CSAM). Specifically, it bans AI systems that produce child sexual abuse imagery or deepfake pornography.

Furthermore, the law prohibits AI systems from engaging in text conversations that simulate sexual content while impersonating children. This targets generative AI models that could be used to create harmful and exploitative interactions.

These provisions reflect a strong stance on protecting children from online exploitation facilitated by AI, aligning with broader efforts to combat child abuse material.

TRAIGA prohibits government entities within Texas from using AI for social scoring. Social scoring typically involves evaluating or classifying individuals based on their social behavior, economic status, or personal characteristics, often leading to discriminatory outcomes or restrictions on rights.

Additionally, the law bans government entities from using AI for biometric identification of specific individuals without their explicit consent. This ensures that biometric data, such as facial recognition or fingerprints, is not used by the government for identification purposes without the individual's knowledge and agreement.

These prohibitions aim to safeguard individual liberties and prevent the misuse of powerful AI technologies by state and local government bodies.

This provision mandates that any Texas government entity making an AI system available to consumers must provide clear notice that consumers are interacting with an AI. This ensures transparency in interactions between the public and automated government services.

The requirement applies specifically when a consumer, defined as an individual acting in a personal or household context, is engaging with a government-deployed AI system. This could include chatbots on government websites, automated phone systems, or other AI-powered interfaces.

The goal is to inform individuals when they are not communicating with a human, allowing them to adjust their expectations and interactions accordingly.

Enforcement authority for TRAIGA rests exclusively with the Texas Attorney General (AG). The AG is empowered to investigate alleged violations of the Act and impose civil penalties.

Before initiating any enforcement action, the AG is required to provide written notice to the alleged violator. This notice triggers a 60-day period during which the entity can respond to the allegations. This pre-enforcement procedure offers a crucial opportunity for remediation and dialogue before formal legal action is taken.

While the Act grants the AG the power to impose civil penalties, the specific amounts of these penalties are not detailed within TRAIGA itself. This leaves the determination of penalty amounts to the AG's discretion and potential judicial review.

TRAIGA includes important safe harbor provisions that offer affirmative defenses against enforcement actions. These provisions incentivize proactive compliance and responsible AI practices.

There are three main circumstances for safe harbor: first, when third parties misuse an AI system in ways prohibited by TRAIGA, protecting the original developer/deployer from liability. Second, if violations are discovered through good faith testing or audits, demonstrating a commitment to self-correction. Third, substantial compliance with the NIST AI Risk Management Framework (AI RMF) or similar recognized standards provides a defense.

These safe harbors are particularly significant for platform operators and developers, as they offer protection when users or external factors lead to non-compliance, provided the entity has taken reasonable steps to ensure responsible use and adherence to recognized standards.

While TRAIGA provides general AI governance, companion legislation, Senate Bill 1188 (SB 1188), specifically addresses the integration of AI in healthcare within Texas. This bill focuses on AI systems used in medical diagnosis.

SB 1188 mandates that licensed practitioners must review AI-generated records. This ensures that human oversight remains central to diagnostic processes, even when AI tools are utilized. The oversight and standards for this review fall under the jurisdiction of the Texas Medical Board.

This sector-specific legislation highlights the critical nature of AI applications in healthcare and ensures that existing medical regulatory bodies maintain authority over the use of AI in patient care.