A Lesson in Legal Realism: The (trimmed down) Texas Responsible AI Governance Act
The Texas Responsible AI Governance Act (“TRAIGA”) stands as one of the most closely watched artificial intelligence bills in the United States. Introduced by Texas State Representative Giovanni Capriglione (R-District 98) on December 23, 2024, this legislation has undergone significant revisions in its scope and impact, likely in line with the current federal government’s attitudes toward AI regulation. Does this new version, filed in March 2025, stand a chance?
TRAIGA’s Original Vision
As first drafted in late 2024, TRAIGA represented an ambitious and comprehensive approach to AI regulation. The original bill mirrored elements of the Colorado AI Act and the EU AI Act, by establishing substantial obligations for developers, deployers, and distributors of “high-risk AI systems." Since late 2024, however, much has changed. The Trump-Vance Administration has made clear it views most laws seeking to regulate AI as “barriers” to American innovation. Governor Youngkin (R) of Virginia recently vetoed the Virginia AI Act, a proposed law similar to the Colorado AI Act. And some tech companies, such as OpenAI, are asking the federal government to treat private AI as exempt from state laws altogether.
The initial 43-page version of TRAIGA included:
A focus on "high-risk AI systems" defined as those playing a substantive role in consequential life-affecting decisions
Requirements for AI deployers to complete an annual impact assessments and within 90 days after substantial modification for every AI system deployed
Mandatory risk management policies for developers and deployers
Detailed disclosure requirements to consumers interacting with AI systems
Obligations for social media platforms and digital service providers (to post terms of service prohibiting deployment of AI system that violates the TRAIGA rules)
Reporting requirements if deployer discovers AI system caused algorithmic discrimination of an individual or group
A general duty of care requiring reasonable measures to prevent algorithmic discrimination
A ban on AI systems that produce or distribute visual material in violation of Texas law (such as depiction of sexual imagery involving a minor)
Giving consumers a right to opt out of the sale of personal data for use in AI systems prior to that data being collected
For employers specifically, the bill as originally filed would have required:
Human oversight of high-risk AI systems
Prompt reporting of discrimination risks
Regular assessments of AI tools
Suspension of non-compliant systems
Frequent impact assessments
Clear disclosure of AI use to individuals
TRAIGA 2.0: A Different Approach
On March 14, 2025, Representative Capriglione filed a substantially trimmed down version of TRAIGA (now HB 149). This new bill shifts the focus primarily to AI systems developed and/or deployed by government agencies rather than comprehensive regulation of private sector AI systems.
Key changes in the revised 31-page bill include:
Complete removal of the "high-risk AI systems" framework
Primary focus on government-deployed AI systems rather than commercial applications
Elimination of impact assessment requirements
Replacement of the duty of care provisions with a narrower prohibition against intentional discrimination
Addition of new prohibitions against manipulation of human behavior to incite harm
Prohibition of social scoring by government AI only (not private sector)
Simplified disclosure requirements, now applying only to government-deployed AI systems
Removal of risk mitigation policy requirements
Elimination of obligations for social media platforms
Removal of reporting requirements for algorithmic discrimination
Ban on AI systems only if developed or deployed “with the sole intent” of producing visual material prohibited by Texas law
Removal of consumer opt-out provisions
Key Differences Between Versions
(Pictured: A table highlighting the key differences between the former and current visions of TRAIGA)
The revised bill appears to be guided by a different regulatory philosophy, perhaps mirroring the federal government’s approach to AI regulation. While the original bill took a proactive, preventative approach, including by requiring businesses to demonstrate compliance through assessments and documentation, the new version does away with most applications to private sector AI.
Other major differences include:
Scope: Original version broadly covered private and public sectors; new version primarily regulates government AI with limited private sector provisions.
Approach to Discrimination: Original version required proactive measures to prevent algorithmic discrimination; new version only prohibits AI developed "with the intent to unlawfully discriminate" and explicitly states that "disparate impact alone does not suffice to show intent to discriminate."
Disclosure Requirements: Original version contained extensive disclosure obligations; new version only requires government agencies to disclose AI use to consumers.
Impact Assessments: Original version required regular assessments; new version eliminates this requirement entirely.
Biometric Identifiers: Original version prohibited any entity from deploying AI using biometric identifiers to identify individuals; new version only prohibits government entities from doing so.
Current Requirements Under TRAIGA 2.0 (IF PASSED)
If passed in its current form, TRAIGA would:
Prohibit the development or deployment of AI "with the intent to unlawfully discriminate against a protected class"
Ban AI systems intentionally designed to encourage self-harm, harm to others, or criminal activity
Prohibit government use of AI for social scoring
Require government agencies to disclose when consumers interact with AI systems
Ban AI systems developed with the “sole intent" of producing unlawful visual material or deepfakes
Create the Texas Artificial Intelligence Council to ensure AI systems uphold ethical standards and serve the public interest
Establish enforcement through the Texas Attorney General's office with no private right of action
Violations not related to unacceptable uses would incur fines of $10,000 to $12,000 per violation, while violations related to unacceptable uses could result in fines of $80,000 to $200,000.
Next Steps
The Texas Legislature's 89th regular session began on January 14, 2025, and will adjourn on June 2, 2025. Representative Capriglione officially filed the revised TRAIGA (HB 149) on March 14, 2025, and it now progresses through the legislative process.
For the bill to become law, it must pass both the Texas House of Representatives and Senate before Governor Greg Abbott signs it. If passed and signed, TRAIGA would take effect on September 1, 2025.
The bill's sponsor, Representative Capriglione, wields significant influence as the author of Texas's Data Privacy and Security Act and co-chair of the state's AI Advisory Council. He also chairs the Texas DOGE committee, a state-level committee that will seek to reduce the size and scope of government.
