The United States is heading into a historic showdown over who governs artificial intelligence. The White House is pushing a national framework to override state laws. States are pushing back hard. Big Tech is spending hundreds of millions to shape the outcome. And somewhere in the middle, businesses are trying to figure out what the rules actually are.
Introduction: The Biggest Power Struggle in Tech History
In the summer of 2025, the U.S. witnessed something unprecedented: a coalition of 42 state attorneys general sent a joint letter to the major artificial intelligence companies demanding better safeguards for children. Two days later, President Trump signed an executive order aimed not at protecting those children — but at dismantling the state laws those same attorneys general were trying to pass.
Welcome to the AI regulation war.
It is a conflict playing out simultaneously in the halls of Congress, in state legislatures from Sacramento to Albany, in federal courtrooms, and in the back rooms where lobbyists whisper policy language to lawmakers. On one side: a growing number of states convinced that AI poses real, tangible risks to their residents and that Washington can’t — or won’t — act fast enough to address them. On the other side: the White House, backed by some of the most powerful and richest companies in American history, pushing to establish a single national standard that would sweep away the emerging patchwork of state rules.
The stakes couldn’t be higher. The decisions made in this regulatory battle will determine how AI is developed, deployed, and held accountable for decades to come. They will shape whether companies face meaningful oversight when their algorithms deny people loans, reject job applications, or flag health insurance claims. They will determine whether your state’s lawmakers have any say over technology that is reshaping every aspect of American life.
And right now, that war is wide open.
How We Got Here: The Rise of the Regulatory Patchwork
For most of the last decade, the federal government took a largely hands-off approach to AI regulation. There were guidelines, voluntary frameworks, sector-specific agency rules, and occasional congressional hearings — but no comprehensive federal law governing how AI could be used in consequential decisions affecting ordinary Americans.
Nature abhors a vacuum. So do state legislators watching their constituents face opaque algorithmic decisions about their housing, employment, credit, and healthcare.
The result was a wave of state AI legislation unlike anything seen before. In 2025 alone, all 50 states introduced some form of AI-related legislation, according to the National Conference of State Legislatures. Some bills passed; many didn’t. But the trajectory was unmistakable: states were going to regulate AI whether Washington wanted them to or not.
Colorado led the way most dramatically. In May 2024, Governor Jared Polis signed SB 24-205, what many called the first comprehensive AI consumer protection law in the United States — one that required companies using high-risk AI systems to conduct algorithmic impact assessments, disclose AI use to affected consumers, and protect against algorithmic discrimination in consequential decisions. The law was modeled partly on the European Union’s AI Act and was hailed by consumer advocates as a national template.
Industry hated it. Tech lobbying groups called the requirements “unworkable.” Governor Polis himself signed the bill with reservations, publicly asking the legislature to revisit it. Compliance costs were flagged as prohibitive. And throughout 2025, industry groups lobbied aggressively to gut the law before it could take effect. When an amendment bill failed in mid-2025, industry shifted strategies — turning their energy toward federal preemption as the ultimate solution.
Other states weren’t far behind. California, Utah, Texas, Connecticut, and New York all advanced AI legislation during this period, each with different approaches, different scopes, and different enforcement mechanisms. For a company operating nationally, the compliance picture was becoming nightmarish — not because any single law was unreasonable, but because all of them were different.
That patchwork became the central argument of the industry’s lobbying campaign: not that AI shouldn’t be regulated, but that it must be regulated consistently, at the federal level, with a single set of rules. It was a position tailor-made for the incoming Trump administration.
The White House’s Power Move: The “One Rule” Strategy
On December 11, 2025, President Trump signed Executive Order 14365: “Ensuring a National Policy Framework for Artificial Intelligence.” The order sent shockwaves through every state capitol that had been working on AI legislation.
The executive order established what critics immediately dubbed the “One Rule” strategy — a coordinated federal campaign to displace state AI laws through a combination of litigation, regulatory reinterpretation, and financial coercion.
The order’s key weapons:
The AI Litigation Task Force. The Justice Department was directed to establish a dedicated task force — within 30 days — whose sole mandate would be to challenge state AI laws in federal court on constitutional grounds. The grounds cited included the Dormant Commerce Clause (the argument that state laws unconstitutionally burden interstate commerce) and conflict preemption (the argument that state rules are incompatible with federal law). It was an aggressive, unprecedented move: a presidential directive telling DOJ to systematically attack laws passed by democratically elected state legislatures.
Federal Funding as a Weapon. The order authorized federal agencies to condition discretionary grants on states agreeing not to enforce AI laws deemed inconsistent with White House policy. Most dramatically, it instructed the Commerce Department to condition $42 billion in previously allocated broadband infrastructure funding — already promised to states under the BEAD program — on the repeal of what the administration called “onerous” AI regulations. For cash-strapped states, this was an existential threat.
FTC and FCC Preemption Plays. The FTC was directed to issue a policy statement characterizing certain state-mandated AI bias mitigation requirements as “per se deceptive trade practices” under the FTC Act — a creative legal theory designed to create a federal ground for preemption. The FCC was directed to open proceedings on whether to adopt a federal AI reporting standard that would supersede state equivalents.
The executive order was careful to carve out certain categories of state law — child safety protections, AI infrastructure siting, state government procurement — from preemption efforts. But its intent was clear: the Trump administration would use every lever of executive power to prevent a state-by-state AI regulatory regime from taking hold.
Three months later, on March 20, 2026, the White House released its National Policy Framework for Artificial Intelligence — essentially a legislative roadmap presented to Congress for enacting the administration’s vision into binding law. The framework called for broad federal preemption of state AI laws that impose “undue burdens,” limitations on state ability to regulate AI model development, restrictions on holding AI developers liable for third-party misuse, and creation of regulatory “sandboxes” to encourage AI experimentation.
Senator Marsha Blackburn introduced the TRUMP AMERICA AI Act — the Republic Unifying Meritocratic Performance Advancing Machine Intelligence by Eliminating Regulatory Interstate Chaos Act — to operationalize the framework. The acronym tells you a great deal about the political environment in which this debate is taking place.
Big Tech’s Billion-Dollar Influence Machine
Understanding this regulatory battle requires understanding the money flowing through it.
In the first three months of 2026 alone, 11 top tech companies — including Alphabet, Microsoft, Anthropic, and OpenAI — spent $20 million on federal lobbying. That’s an average of $226,000 per day, according to an analysis of Q1 2026 lobbying reports by the bipartisan reform group Issue One. Big Tech’s lobbying expenditures have nearly doubled since 2020.
Meta leads the pack, spending $7.1 million — nearly $80,000 a day — on federal lobbying in just the first quarter of 2026. Anthropic quadrupled its congressional lobbying year-over-year, reaching $1.56 million in Q1, compared to $360,000 the previous year. OpenAI nearly doubled its federal lobbying, rising from $560,000 to $1.02 million in the same period. All told, Alphabet, Meta, Microsoft, Nvidia, Anthropic, and OpenAI employed 307 registered lobbyists during that quarter alone.
State-level spending is even more revealing. In California alone, AI and tech companies invested more than $39 million to influence state politics in 2025. Meta spent $4.6 million lobbying California state officials in a single year — the highest in that company’s history of state-level advocacy in Sacramento. Google spent more than $3.5 million lobbying on AI-related issues. When combined with the $1.1 billion in total tech political spending analyzed by the consumer advocacy group Public Citizen across the 2024-2025 cycle, the scale of industry influence becomes staggering.
But the tech industry isn’t speaking with one voice — and that internal fracture is one of the most interesting dynamics in this debate.
OpenAI and Microsoft have pushed for a federal licensing regime, arguing that the most powerful AI models pose risks comparable to nuclear weapons or pandemics, and that such existential risks belong exclusively to federal jurisdiction. Critics of this position argue it’s a sophisticated form of regulatory capture — erecting high barriers to entry that would protect the market dominance of incumbents while making it prohibitively expensive for smaller competitors to operate.
Meta and Andreessen Horowitz take the opposite position within the tech world: they oppose any framework that imposes liability on developers of “open weights” models. Their lobbying is less about controlling a licensing regime and more about protecting the open-source AI ecosystem from what they see as existential legal threats.
Other voices — including academic researchers, civil rights organizations, and open-source AI advocates — worry that a Big Tech-authored federal preemption framework would be the worst possible outcome: a national standard written by the companies it’s supposed to govern, eliminating the messy-but-democratic state-level experimentation that has produced many of the consumer protections currently on the books.
The States Fight Back
The executive order has not gone unopposed. Far from it.
Congress has repeatedly refused to enact the federal preemption the White House has sought. Attempts to insert a 10-year moratorium on state AI regulations into the National Defense Authorization Act for 2026 failed after bipartisan opposition, including from House Armed Services Committee Chairman Mike Rogers. The moratorium also failed to survive the One Big Beautiful Bill Act. Despite executive pressure, the legislative path to comprehensive federal preemption remains uncertain.
Democrats have organized a counter-offensive. Representative Doris Matsui and colleagues introduced the GUARDRAILS Act on March 20, 2026 — the same day the White House released its framework — which would repeal the December executive order and block federal preemption of state AI regulation. Senator Schatz has introduced companion legislation in the Senate. The political battle lines are clear: Republicans generally support federal preemption; Democrats generally oppose it, arguing states have a fundamental right to protect their residents.
The constitutional challenges to the executive order’s preemption theories are substantial. As legal experts at multiple major firms have noted, executive orders cannot independently displace state laws — that generally requires an act of Congress. The FTC’s ability to preempt state bias-mitigation laws through a policy statement faces serious questions about statutory authority. Courts have not yet delivered definitive rulings on these questions, but the litigation is coming.
Meanwhile, states are adapting. Colorado ultimately rewrote its entire AI law — Governor Polis signed the replacement bill, SB 26-189, on May 14, 2026 — scaling back the original’s broad governance requirements in favor of a more targeted disclosure-and-transparency framework. The revision was partly a response to industry pressure, but the law survived. It goes into effect January 1, 2027. California, Texas, Utah, and New York have continued advancing their own frameworks, creating exactly the patchwork the White House claims to be trying to eliminate.
There’s also a growing coalition of state officials actively resisting federal pressure. The 42-state attorney general coalition that wrote to AI companies in late 2025 is not going away quietly. State AGs have independent enforcement authority and constitutional standing to defend their own laws. For the White House’s “One Rule” strategy to fully succeed, it would need to win in court against determined opposition from nearly every major state in the country.
What This Means for Businesses: Navigating Uncertainty
For companies deploying AI in consequential decisions — lenders, insurers, healthcare organizations, employers, landlords — the regulatory uncertainty created by this battle is its own kind of cost.
The most prudent legal guidance from across the spectrum is consistent: do not assume state laws will be preempted in the short term. Congress has not passed a federal AI law. Executive orders alone cannot preempt state statutes. The litigation will take years to resolve. In the meantime, Colorado’s revised law, California’s transparency requirements, Texas’s biometric data rules, and New York City’s automated employment decision regulations are all real, enforceable obligations.
That said, companies with operations in multiple states face genuine compliance complexity. A loan algorithm that satisfies Colorado’s explanation requirements may need to be separately audited under California’s privacy regulations and yet again calibrated against New York’s employment rules. For larger enterprises with sophisticated legal and compliance teams, this is manageable — expensive, but manageable. For smaller companies and mid-market players, the patchwork is a genuine operational burden.
The companies best positioned in this environment are those building AI governance infrastructure regardless of which regulatory framework ultimately prevails. The requirements that keep appearing across state laws — explainability, human review of adverse decisions, consumer disclosure, bias testing, record-keeping — are not going away no matter what happens in Washington. Whether mandated by Colorado’s AG or a future federal standard, these capabilities represent table stakes for responsible AI deployment.
The Global Context: America Is Already Behind
Lost in the domestic political noise is a broader competitive reality: while the United States debates whether to regulate AI at all, the rest of the world has moved forward.
The European Union’s AI Act is in force, imposing tiered requirements on AI systems based on risk level, with strict obligations for high-risk applications in employment, healthcare, credit, and education. Canada has advanced the Artificial Intelligence and Data Act as part of Bill C-27. The United Kingdom, Australia, Singapore, Japan, and South Korea all have active AI governance frameworks in development or implementation.
The White House argues that excessive regulation will cede AI leadership to China. The counter-argument — made by consumer advocates, civil rights organizations, and many state officials — is that a race to the bottom on AI governance will ultimately undermine public trust in American AI systems, creating a different kind of competitive disadvantage as global customers demand responsible, explainable, auditable AI.
The tension between innovation speed and governance rigor is real. But the framing of “regulation vs. innovation” misrepresents how the most sophisticated companies actually operate. Companies selling AI-powered financial services into Europe already comply with the EU AI Act. Companies operating healthcare AI in multiple countries already build explainability into their systems. The marginal cost of complying with thoughtful U.S. state laws is far lower than industry lobbying campaigns suggest — particularly compared to the cost of a major enforcement action, a class-action lawsuit under existing civil rights law, or the reputational damage of an AI discrimination scandal.
Who Will Win?
The honest answer is: nobody knows yet, and the outcome will almost certainly be a compromise that satisfies nobody completely.
A full federal preemption of all state AI laws is politically and constitutionally unlikely. Democrats and state-rights Republicans have repeatedly blocked it in Congress. Courts are skeptical of executive-order-based preemption. And the political cost of appearing to protect Big Tech at the expense of consumers is significant in an election year.
A return to complete state-by-state fragmentation is also unlikely. Industry’s compliance cost arguments, while often overstated, have some legitimate basis. And there’s a reasonable federalism argument that some aspects of AI governance — particularly around foundational model development, interstate commerce, and national security — genuinely belong at the federal level.
The most probable outcome is a messy middle: a federal framework that sets baseline standards in specific sectors (financial services, healthcare, employment), preempts the most onerous state provisions in those sectors, but preserves state authority on consumer protection, civil rights enforcement, and child safety. That’s roughly the shape of how federal-state regulatory coexistence works in financial services, privacy law, and environmental regulation today.
What’s different with AI is the pace. The technology is moving faster than the legal system can respond, the political pressures are more intense than usual regulatory battles, and the stakes — for innovation, for equity, for democratic accountability — are higher than almost any governance question in recent American history.
Conclusion: Why This Fight Matters to Everyone
The AI regulation war isn’t just a story about lobbyists, legislators, and legal theories. It’s a story about power — about who gets to decide the rules governing systems that are increasingly making decisions about people’s jobs, homes, credit, and healthcare.
The companies arguing for minimal federal oversight have a legitimate interest in operational predictability. The states arguing for their right to protect residents have a legitimate interest in democratic accountability. The individuals whose lives are shaped by algorithmic decisions have a legitimate interest in systems that are fair, transparent, and subject to meaningful challenge.
None of those interests is entirely wrong. The question is how to weigh them — and who gets to weigh them.
For now, the answer is: everyone is fighting it out simultaneously in every arena available, with more money and political capital than has ever been deployed in a technology policy battle. Businesses navigating this environment cannot afford to wait for clarity. The companies that build governance infrastructure now — explainability capabilities, bias auditing, consumer disclosure workflows, human review processes — will be ready for whatever regulatory framework eventually emerges.
Because one thing is certain: AI will be regulated. The only question is by whom, on what terms, and at whose expense.
Sources: Paul Hastings LLP (December 2025); Latham & Watkins (December 2025); WilmerHale (March 2026); Holland & Knight (March 2026); Ropes & Gray (March 2026); Akin Gump (March 2026); Fortune / Issue One (April 2026); CalMatters (March 2026); Public Citizen (November 2025); GovFacts (December 2025); Colorado SB 26-189 (signed May 14, 2026).