Texas Just Created A New Model for State AI Regulation
Austin, Texas—Texas Governor Greg Abbott (R) last month signed into law the Texas Responsible Artificial Intelligence Governance Act (TRAIGA), joining Colorado as only the second state to adopt a comprehensive AI governance law. This new legislation marks a significant shift in the landscape of artificial intelligence regulation.
A New Standard for AI Regulation?
TRAIGA introduces a novel approach to AI regulation by limiting the state’s ability to punish companies while expanding investigatory powers. When implemented on January 1, 2026, the law is expected to subject many private enterprises to regulatory review but few to actual punishment. Given Texas’ substantial economic influence, the law is likely to have wide-ranging implications for the AI industry.
The backdrop for this legislation is the recent demise of the state AI moratorium provision in the federal budget reconciliation bill, which indicates that AI-related lawmaking will continue to evolve. Some jurisdictions, like Colorado, are reconsidering comprehensive AI laws, while others are moving towards sector-specific regulations.
Key Features of TRAIGA
TRAIGA forges a middle path, focusing on AI usage across both public and private sectors rather than the inherent power of AI models. Key features of TRAIGA include:
- Limited Enforcement: Enforcement is restricted to a small number of enumerated harms.
- Affirmative Defenses: Violators have opportunities to cure defects.
- Biometric Data Limits: Some restrictions on the collection of biometric data.
- Broad Investigatory Authority: The attorney general is vested with extensive powers to investigate.
TRAIGA also establishes a Sandbox Program that allows companies to test AI systems without full regulatory compliance, although core prohibitions still apply.
Pared-Back but Still Far-Reaching
Initially introduced in December 2024, TRAIGA underwent significant modifications before its enactment. Despite being narrower than its original proposal, TRAIGA imposes considerable regulatory burdens on a wide range of private entities due to:
- Broad Jurisdiction: The law applies to any person or entity conducting business in Texas or developing AI systems used by Texans.
- Expansive Definition of AI: TRAIGA’s definition encompasses any machine-based system that infers outputs from inputs.
- Inclusion of Developers and Deployers: It applies to both developers and deployers, covering a wide array of companies.
Prohibitions and Duties
TRAIGA imposes specific prohibitions on AI developers and deployers, including:
- Prohibition against AI systems that intentionally incite self-harm or criminal activity.
- Restrictions on infringing individual rights under the US Constitution.
- Ban on unlawful discrimination against protected classes.
- Prohibition against producing or distributing certain sexually explicit content.
Furthermore, Texas government agencies are mandated to disclose when citizens interact with AI systems and are restricted from using AI for social scoring.
Investigation and Enforcement
The Texas attorney general holds exclusive enforcement authority under TRAIGA. The law allows consumers to submit complaints about AI systems, enabling the attorney general to issue civil investigative demands (CIDs) based on a single complaint, even without establishing probable cause.
This broad authority means that if a Texan alleges that a large language model censors speech, both the model’s developer and deployer could be subjected to investigation.
Opportunity to Cure
While TRAIGA facilitates investigations, it complicates penalties for wrongdoers. Violators have 60 days to rectify violations after receiving notice from the attorney general. They may also assert affirmative defenses if they demonstrate compliance with recognized risk management frameworks.
Penalties
The attorney general can seek penalties reaching up to $12,000 for each curable violation and $200,000 for uncurable ones, along with additional penalties for state-registered entities.
Regulatory Sandbox Program
TRAIGA’s Sandbox Program promotes AI use in sectors like healthcare and finance while prohibiting charges against participants. However, participants are still bound by TRAIGA’s core prohibitions.
Conclusion
TRAIGA positions Texas as a potential leader in AI governance, setting a new standard for how states can regulate this rapidly evolving field. By balancing innovation with necessary protections, Texas aims to facilitate responsible AI development while safeguarding individual rights.
