What it is

House Bill 149 — the Texas Responsible Artificial Intelligence Governance Act. Signed by Governor Greg Abbott on June 22, 2025. Effective January 1, 2026.

The final bill is substantially narrower than the original draft. Most of the broad “high-risk AI” obligations modeled on the EU AI Act were stripped before passage; what remains is a targeted prohibition list, a sandbox program, and an advisory council.

What it prohibits

Development or deployment of AI systems that are designed to:

  • Incite or encourage self-harm, harm to others, or criminal activity
  • Produce child sexual abuse material or non-consensual deepfake intimate imagery
  • Engage in unlawful discrimination (state-action context)
  • Infringe constitutional rights

Government entities are separately prohibited from using AI for “social scoring” of individuals.

What it creates

  • Texas Regulatory Sandbox Program — a 36-month testing pathway for AI developers, administered by the Department of Information Resources, with limited liability for participants
  • Texas Artificial Intelligence Advisory Council — a state body charged with monitoring AI use across government and recommending legislative refinements

Enforcement

  • Exclusive enforcement authority sits with the Texas Attorney General. No private right of action.
  • Cure period — 60 days for violators to remediate before penalties attach.
  • Civil penalties — $10,000 to $200,000 per violation (depending on whether the conduct is curable), or $2,000 to $40,000 per day for continuing violations.
  • Disgorgement — the AG may seek injunctive relief and recovery of attorney’s fees.

Who it covers

Persons who develop, sell, or deploy AI systems in Texas, or whose AI systems’ output is used in Texas. The narrowed scope makes most general-purpose enterprise AI use outside the prohibited categories, but in-scope conduct (especially around content generation and government use) is now subject to live enforcement.

Practical posture

  • Map any generative AI deployment against the prohibited-use list, particularly around content moderation, image generation, and chatbot guardrails
  • Document policy controls and incident-response procedures — TRAIGA’s curable/non-curable distinction makes evidence of remediation directly load-bearing on penalty exposure
  • Government contractors should expect contract clauses referencing TRAIGA’s social-scoring prohibition
  • Developers exploring novel AI applications should evaluate the sandbox program; it is one of the more favorable state-level testing regimes in the US

What to watch

  • AG enforcement priorities and first published actions
  • Sandbox program rules issued by the Department of Information Resources
  • Recommendations from the Advisory Council that may seed a TRAIGA 2.0