Texas Responsible Artificial Intelligence Governance Act (TRAIGA)

Intro

This page presents the current version (v1.1) of a targeted Phase II add-on to the Texas Responsible Artificial Intelligence Governance Act (HB 149). While HB 149 establishes statewide rules governing artificial intelligence systems—including prohibited uses, transparency requirements, enforcement authority, and a regulatory sandbox—this proposal addresses a specific remaining gap: artificial intelligence systems that are configured to initiate or execute real-world actions in high-stakes contexts. The add-on applies only at deployment, preserves innovation by excluding research, model training, and open-source development, and uses existing enforcement structures. Its purpose is to ensure that when artificial intelligence systems are permitted to act with real-world consequences, those actions remain subject to meaningful human authorization, auditability, and fail-safe safeguards.

The Problem

Artificial intelligence systems are increasingly deployed in areas with significant consequences for Texans, including healthcare, employment, financial services, housing access, public benefits, infrastructure, and core government operations.

House Bill 149 (TRAIGA) establishes important statewide protections and governance mechanisms for artificial intelligence. However, it does not explicitly address agentic AI systems—systems that are configured to initiate or execute actions without contemporaneous human approval.

In high-stakes settings, this creates an accountability gap:
who is responsible when an automated system acts, errs, or causes harm?

The Solution

This proposal adds a narrow deployment safeguard to TRAIGA that applies only when AI systems are permitted to take or trigger high-stakes actions.

The add-on requires:

Meaningful human authorization before consequential actions

• Audit logs and traceability for actions and overrides

• Fail-safe defaults when systems encounter uncertainty or anomalies

• Ongoing monitoring and incident response

• Human review and explanation rights for affected individuals

The focus is on how AI is deployed, not how it is built.

What This Proposal Does Not Do

This add-on is intentionally limited in scope. It does not:

• Ban artificial intelligence

• Regulate AI research, training, or compute

• Restrict open-source models

• Create a new AI agency or bureaucracy

• Regulate speech or content moderation

It applies only to deployment-level use of high-stakes, action-taking systems.

How This Helps Legislative Offices

• Provides clear amendment-ready legislative language

• Supports committee preparation and testimony

• Lays out statutory context for efficient legal review

• Offers ready-to-use materials for stakeholder engagement

Prepared By

Kevin Bihan-Poudec
Dallas, Texas (75219)
Voice For Change Foundation

info@voiceforchangefoundation.org
www.voiceforchangefoundation.org