Estimated Time to Read: 8 minutes
Artificial intelligence (AI) is already embedded in healthcare, housing, finance, employment, and law enforcement. As its adoption spreads, state legislatures are competing to set the rules of the road. In New York, lawmakers have proposed mandatory audits for bias, Colorado has passed annual reporting requirements, and California has advanced disclosure mandates that could force the release of sensitive training data. Illinois wants disclosures for all AI interactions, while Maine requires them only in cases of possible deception.
Texas, too, has entered the debate. During the 88th Legislature, then-House Speaker Dade Phelan (R-Beaumont) created the House Select Committee on Artificial Intelligence and Emerging Technologies to study the issue during the interim and report findings ahead of the 89th Legislative Session (2025). That interim report, along with several major bills filed once the Legislature convened, reflects the same tension seen nationally: a desire to protect consumers and safeguard civil liberties, but at the cost of expanding bureaucracy, increasing taxpayer burdens, and risking innovation.
The broader challenge is that without federal preemption, America faces a regulatory maze in which states move in competing directions. This patchwork not only drives up compliance costs but also allows individual states to set de facto national standards, undermining predictability for innovators and clarity for consumers.
National Trends in AI Legislation
The Future of Privacy Forum’s State AI Legislation Report highlights common trends across state efforts. Most focus on so-called “consequential decisions” in critical areas like education, employment, housing, healthcare, and finance. Some states, like California, opt for outright prohibitions on discriminatory AI use. Others, such as Colorado, adopt a duty of care model, requiring developers and deployers to take reasonable steps to avoid bias.
The report also shows that lawmakers are drawing distinctions between developers and deployers. Developers are expected to test systems, provide documentation, and disclose risks, while deployers must monitor outcomes, notify consumers, and conduct risk assessments. Consumer rights are also growing, with provisions granting notice, explanation, correction, and appeal rights to individuals impacted by AI-driven decisions. Enforcement often rests with state attorneys general, who may investigate violations and impose penalties.
Alongside these frameworks, states are experimenting with rules for generative AI and frontier models. Some now require content labeling, watermarking, or disclosure of training data. Others have left these technologies largely unregulated. The result is an inconsistent and unpredictable landscape that complicates compliance, stifles smaller competitors, and leaves consumers with uneven protections.
Why Federal Preemption Matters
The case for federal preemption is straightforward. Fragmentation drives companies to spend more time navigating compliance paperwork than addressing real risks. Large corporations with armies of lawyers can adapt, while smaller firms are discouraged from innovating. Consumers face confusion, with rights and protections depending on where they live.
A federal framework would harmonize obligations, provide predictability, and ensure consumer rights apply consistently nationwide. By reducing duplicative compliance costs and preventing any one state from setting de facto national standards, federal preemption would promote both innovation and fairness. Without it, America risks entrenching incumbents and undermining free enterprise.
Texas as a Case Study in State-Level AI Regulation
The 89th Texas Legislature illustrates how quickly well-intentioned proposals can spiral into bureaucratic growth and fiscal burdens.
House Bill 149 (HB 149), authored by State Rep. Giovanni Capriglione (R-Southlake), sought to create a comprehensive AI regulatory regime, including oversight of biometric data, developer and deployer mandates, and the creation of the Texas Artificial Intelligence Council. Even after revisions, the bill carried a projected cost of over $25 million for the biennium and required twenty new state employees. Its vague ethical mandates and compliance obligations risked chilling innovation among small businesses and open-source developers. Texas Policy Research recommended opposing HB 149 unless it was amended with strict limits, sunset provisions, and small business exemptions. HB 149 passed overwhelmingly and fully goes into effect on January 1st, 2026.
Senate Bill 1964 (SB 1964), authored by State Sen. Tan Parker (R-Flower Mound), targeted government use of AI by requiring system inventories, classifying “heightened scrutiny” tools, and mandating adherence to an AI Code of Ethics. The substitute version removed important transparency features like public complaint portals, leaving only bureaucratic expansion. With a projected biennial cost of $7.28 million and recurring annual expenses of over $4 million, the Department of Information Resources would gain new powers and staff at taxpayers’ expense. Vendors contracting with the state would face new compliance risks that could discourage small businesses from participating. SB 1964 passed easily and has already gone into effect.
House Bill 2818 (HB 2818), also authored by Capriglione, proposed creating a new AI Division within the Department of Information Resources to modernize legacy systems with generative AI. Although it aimed at efficiency, the bill imposed an $8.1 million cost through 2027 and nearly $5 million in annual recurring costs thereafter. It authorized nine new employees but included no sunset provisions, benchmarks, or budgetary limits. While it did not regulate private firms directly, it risked duplicating private-sector functions and expanding state infrastructure without accountability.
Taken together, these bills reveal a pattern: Texas lawmakers are following national trends in AI regulation but doing so in ways that expand bureaucracy and impose heavy costs, inconsistent with the state’s reputation as a pro-innovation environment. HB 2818 passed easily and has already gone into effect.
Interim Report Findings: Texas’ Legislative Vision
The Texas House Select Committee on Artificial Intelligence and Emerging Technologies, created during the 88th Legislature, released its interim report in May 2024. It explored AI’s implications for homeland security, elections, law enforcement, and cybersecurity. Witnesses highlighted threats posed by deepfakes in political campaigns, the use of AI in identifying child exploitation material, and vulnerabilities in critical infrastructure.
The committee’s recommendations included banning deepfakes in political advertising, updating laws to address AI-generated content, increasing workforce education, and encouraging collaboration between government, academia, and industry. Then-Speaker Dade Phelan emphasized the importance of balancing AI’s opportunities with safeguards against privacy violations and disinformation, promising that these findings would shape the 89th Session.
This report demonstrates that Texas is not only considering sector-specific regulation but is also willing to expand government oversight broadly in the name of addressing AI risks.
The AI Action Plan: Deregulation With Risks
At the federal level, the AI Action Plan released in July 2025 sets out a vision built on three pillars: accelerating innovation, building infrastructure, and leading in international diplomacy. The plan calls for removing red tape, encouraging open-source and open-weight models, streamlining permits for data centers and semiconductor facilities, and expanding grid capacity to power AI.
This deregulatory thrust is the right instinct. It rejects Europe’s prescriptive AI Act and emphasizes America’s strength in permissionless innovation. Encouraging open-source models, protecting free speech, and removing ideological mandates help keep the U.S. competitive.
However, the administration undermines its message by considering equity stakes in Intel and other chipmakers that received CHIPS Act subsidies. What began as a subsidy program has drifted toward creeping nationalization. Government ownership distorts markets, politicizes corporate strategy, and invites lobbying arms races. Intel has already received billions in subsidies; turning those grants into equity stakes amounts to corporate socialism.
The contradiction is striking. Washington cannot credibly claim to champion innovation while embedding itself in corporate balance sheets. If Intel cannot compete without taxpayer equity, the problem is with its business model, not the absence of government ownership.
Liberty Principles and the Bigger Picture
Evaluated against the five liberty principles, both Texas and Washington fall short. Provisions that protect biometric data or ban deepfake political advertising respect individual liberty, yet vague ethical mandates and sweeping investigatory powers threaten speech and innovation. Developers should be responsible for the impacts of their tools, but concentrating oversight in unelected councils undermines personal responsibility. Free enterprise is stifled when small businesses are saddled with compliance burdens while large firms benefit from subsidies and government favoritism. Property rights are strengthened when individuals retain control over biometric data, but weakened when the state inserts itself as a corporate shareholder. And limited government, the principle most consistently ignored, is eroded both by Texas’ new councils and divisions and by Washington’s growing industrial policy.
Conclusion: Trust Markets Over Ministries
Artificial intelligence holds enormous promise, but America’s current approach risks squandering it. State patchworks create conflicting compliance regimes, while federal subsidies drift toward corporate socialism. Texas provides a vivid case study of how quickly enthusiasm for regulation can turn into bureaucratic growth and taxpayer burdens. Washington, meanwhile, undercuts its own deregulatory Action Plan with government ownership schemes that entrench favoritism.
The path forward is clear. Reject fragmented state laws. Resist industrial policy that replaces markets with ministries. Embrace a federal framework grounded in liberty that prioritizes permissionless innovation, consistent rules, and competitive markets. America will win the AI race not by copying Europe’s regulatory excesses or China’s central planning, but by out-freedoming both.
Texas Policy Research relies on the support of generous donors across Texas.
If you found this information helpful, please consider supporting our efforts! Thank you!
