Federal AI Preemption Is Coming. Here's What Texas Keeps and What It Loses.

The framework protects state authority in three areas. Which ones survive, and which ones don't, is the difference between a compliance strategy that'll hold up and one that gets rewritten in 18 months.

Zoning. States keep full control over where AI infrastructure gets built. Every county commission, every water district, every municipal planning board keeps its current authority. For datacenter developers navigating SB6's interconnection requirements, nothing changes on the siting front.

General applicability laws. Texas consumer protection, fraud prevention, and child safety statutes stay enforceable against AI developers and users. The AG's office doesn't lose jurisdiction.

State procurement and services. Texas controls how its own agencies use AI in law enforcement, public education, and government services. DIR's AI Regulatory Sandbox and the Code of Ethics for state AI systems aren't going anywhere.

The framework bars states from three things:

Regulating AI development. The White House calls AI development "an inherently interstate phenomenon with key foreign policy and national security implications." States can't impose requirements on how AI models are built, trained, or tested.

Penalizing developers for third-party misuse. If someone uses an AI model to commit fraud, the state can prosecute the person. It can't hold the model developer liable for that third party's conduct.

Burdening lawful AI use. If an activity is legal without AI, states can't create additional restrictions just because AI is involved.

Texas passed the Responsible AI Governance Act (TRAIGA, HB 149/SB 1964) effective January 1, 2026. It's one of the most carefully calibrated state AI laws in the country. Intent-based standard, not disparate impact. AG-only enforcement, no private right of action. Sixty-day cure period before fines kick in.

Most of TRAIGA likely survives federal preemption. Its core provisions target AI deployment, not development. The enforcement mechanisms (AG action, no private suits) are conservative by design. And the intent standard lines up with where the White House wants to go.

But the 90th Legislature had plans to expand it. Lt. Gov. Dan Patrick's interim charges include grid resilience provisions for AI datacenter power demand and stricter deepfake liability standards. If Congress passes a preemption bill before the 90th convenes in January 2027, some of those planned expansions won't survive contact with federal boundaries.

The White House framework includes a "Ratepayer Protection Pledge" telling Congress to make sure residential electricity customers don't bear increased costs from datacenter construction. It also calls for faster permitting for on-site power generation. Texas already did both through SB6 (signed June 2025).

SB6 forces datacenter developers to pay 100% of interconnection construction costs. $50K/MW in financial security at the intermediate agreement. Another $50K/MW as a non-refundable interconnection fee. No ratepayer subsidies. The PUCT's rulemaking (Project 58481) adds batch study processes, 5-year clawback provisions, and remote disconnection authority for ERCOT.

The White House framework validates the "pay your own way" model Texas built a year ago. It doesn't create new burdens for developers already navigating SB6. It tells 49 other states to follow the path Texas cleared.

Three implications for Texas policymakers and the companies watching them:

The White House framework isn't anti-Texas. It's largely pro-Texas. SB6's cost model, TRAIGA's intent standard, and DIR's sandbox all align with where the Administration wants Congress to go.

The risk is in the details Congress writes. "Preempt state AI laws that impose undue burdens" is a principle, not a statute. Who defines "undue"? Which laws qualify? Those questions will get answered in committee markups and floor amendments over the next 12 to 18 months.

Companies operating in Texas should be tracking three things: Commerce and Judiciary committee responses, the specific preemption language in any draft bills, and how the 90th Legislature positions its AI agenda relative to the federal timeline. Early visibility into these developments is the difference between shaping your compliance strategy and reacting to someone else's.