Fifty Labs, No Standards
What happened
Across the United States, state legislatures are advancing more than 1,500 AI-related bills in 2026, covering liability for AI-generated harm, algorithmic discrimination, deepfake regulation, and required disclosures in hiring and lending. President Trump has explicitly stated his opposition to state-level AI laws, arguing that only a federal framework should govern AI to avoid impeding US competitiveness. No federal AI bill has passed Congress. Simultaneously, the EU AI Act's compliance deadlines are arriving for companies operating in Europe, and Singapore, Vietnam, and South Korea have each issued new AI governance frameworks. US AI developers face a matrix of conflicting requirements with no coherent center.
The US has created the worst possible regulatory environment for AI development: not permissive enough to be a haven, not coherent enough to provide legal certainty, and not unified enough to compete with the EU's single-framework approach.
The Hidden Bet
State-level AI regulation is primarily a compliance burden
Some state laws are establishing liability frameworks that could actually benefit smaller AI developers by giving them legal clarity that the absence of federal law denies them. The complaint that states are creating a 'patchwork' often comes from large tech companies whose real objection is that any regulation is burden, not that state regulation specifically is worse than federal regulation.
Trump's federal-only preference will eventually produce a federal AI law
Congress has failed to pass major tech legislation for years. The same political conditions that blocked federal privacy law, federal social media law, and federal content moderation law apply to federal AI law. Trump opposing state laws without producing a federal substitute may simply mean AI goes unregulated nationally, which benefits incumbents with regulatory moats.
US AI companies are disadvantaged by this patchwork
The largest US AI companies have teams of lawyers who can navigate 50 different state frameworks. The real burden falls on startups, researchers, and smaller companies who cannot afford compliance infrastructure. The patchwork may be filtering out competition for incumbents while appearing to constrain everyone equally.
The Real Disagreement
The real fork is whether AI regulation should be designed to protect the public now, before harms are fully understood, at the cost of innovation friction, or to enable maximum innovation first, at the cost of harms that may prove difficult to reverse. Every state bill leans toward the first. Trump's federal inaction leans toward the second. Both positions reflect genuine values. The case for action now: AI systems affecting hiring, lending, and healthcare are already causing measurable harm to identifiable people, and waiting for Congress means waiting indefinitely. The case for waiting: most specific AI harms are not unique to AI and are addressed by existing discrimination and consumer protection law. I lean toward targeted intervention on high-stakes uses, discrimination in employment and lending specifically, rather than the catch-all state bills that may criminalize beneficial AI uses in the same stroke.
What No One Is Saying
The 1,500-bill number is a political signal, not a regulatory coherent program. Most of these bills will not pass. The ones that do will be the most politically salient, which often means the most reactive to a specific scare rather than the most well-designed to address actual harm. The real outcome of 'states regulating AI' is not 50 thoughtful frameworks. It is 8-10 poorly drafted laws in large states that developers must treat as the national standard by default.
Who Pays
AI startups and mid-size companies
Now. Each new state law that passes raises the compliance floor.
Compliance with even 10 substantively different state frameworks requires legal infrastructure that is proportionally more expensive for smaller organizations. Large incumbents absorb this cost; smaller competitors cannot.
People harmed by AI in high-stakes decisions (hiring, lending, healthcare)
Already occurring. Will continue until either federal law passes or the Supreme Court preempts state laws.
Without a coherent federal liability framework, legal remedies for AI-caused harm are fragmented and jurisdiction-dependent. Harm in a non-regulating state has no legal redress.
US AI competitiveness vs EU
EU Act enforcement begins ramping through 2026-2027.
The EU AI Act provides a single compliance framework for all EU operations. US companies serving both markets must build compliance twice. European AI companies serving only the EU have simpler compliance requirements.
Scenarios
California sets the floor
California passes a comprehensive AI liability bill that becomes the de facto national standard, as happened with CCPA for privacy. Other states defer to it or adopt similar language. Companies build compliance to California standards for all US operations.
Signal California AI bill passes committee with broad business and consumer group support.
Federal preemption
Congress passes a narrow federal AI bill that explicitly preempts state laws. The bill is written primarily by industry and establishes minimal federal standards. State regulation becomes legally inoperative. The patchwork ends but so does meaningful accountability.
Signal Senate AI subcommittee schedules markup of a preemption-first federal AI bill within 90 days.
Regulatory entropy
No federal law passes. State bills keep advancing and some pass, but enforcement is spotty. Companies develop a strategy of technical compliance for large states and effective non-compliance everywhere else. AI use in high-stakes contexts continues without meaningful accountability.
Signal Two years pass with no federal AI bill and at least five contradictory state laws in effect.
What Would Change This
If Congress passed a federal AI bill within 12 months that included meaningful liability for high-stakes uses, the bottom line about worst-of-all-worlds regulatory environment would need revision. If the EU AI Act compliance proves manageable for US companies without creating the predicted competitive disadvantage, the global competitiveness argument weakens.
Related
Zero Federal AI Laws, 1,561 State Bills, and a White House Framework Nobody Has to Follow
powerThe First State to Say No to AI
powerThe Federal Government Just Declared War on State AI Regulation, and Colorado Folded Before the Battle Started
powerThe EPA Just Deleted the Legal Foundation for Every US Climate Rule. 25 States Are Suing.