I Mapped AI Law in All 50 States. Here Is What the Map Says.

#No Federal Law, Fifty State Answers

There is no federal AI law in the United States. While everyone waits for one, fifty state legislatures have written their own, and the result is the most consequential body of technology regulation in the country that almost no one has read end to end.

So I read it. All of it. I built a dedicated, continuously updated breakdown for every single state, each page carrying the actual bills, statuses, effective dates, and primary-source links, plus an executive analysis of what it means for builders and businesses operating in that state. This piece is the view from above: what the fifty-state map shows once you stand back from it.

#The Headline Numbers

More than 1,700 AI bills have been introduced across all 50 states. Over 34 are signed into law, and more than 40 have passed at least one chamber. But the raw count hides the real story, which is that the states have quietly split into two incompatible philosophies of how to govern artificial intelligence.

#Two Americas of AI Regulation

On one side sit the comprehensive regulators. Colorado started it with the first cross-sector AI Act in 2024, a risk-based framework built around algorithmic discrimination and a developer duty of care, then repealed and reenacted it before it ever took effect. California built the most layered stack in the country, anchored by SB 53, the first US law to impose binding safety and transparency duties on frontier model developers. New York followed with the RAISE Act, and Connecticut finally passed its long-fought comprehensive bill after three sessions of trying.

On the other side sit the innovation-protective states. Texas chose intent-based liability over Colorado's outcome-based model and put the heaviest duties on government rather than industry. Montana went furthest, enacting a Right to Compute Act that makes restricting AI subject to strict scrutiny, the highest legal bar there is. Utah built an AI sandbox and an Office of AI Policy instead of a rulebook. Ohio is debating whether to legislate that AI can never be a legal person.

The gap between these two camps is widening, not closing, and it is the single most important fact for anyone deploying AI across state lines. There is no single American AI compliance posture. There are at least two, and they point in opposite directions. This is the same sovereignty problem I wrote about in the hardware layer teardown: control lives wherever someone else sets the rules, and here the rules differ at every border you cross.

#Where the States Actually Agree

Strip away the philosophical fight and a striking consensus appears underneath. On three specific harms, red and blue states have moved in near lockstep.

AI-generated child sexual abuse material is now criminal in a long list of states, and the laws are written to close the no-real-child defense. Nonconsensual sexual deepfakes carry criminal and civil liability almost everywhere that has acted. And election deepfakes, synthetic media impersonating candidates, are now subject to disclosure mandates in a large and growing bloc of states, from Michigan to Minnesota to Wisconsin.

The newest wave is companion chatbots. In a span of months, California, Nebraska, Oregon, Washington, Hawaii, and Maine have passed or advanced laws requiring AI to disclose that it is not human, with heightened protections for minors and self-harm protocols. If you build conversational AI, this is the obligation arriving fastest, and most of these take effect in 2027, which is your runway.

#What Changed Most Recently

The map moved even while I was building it. Connecticut signed its comprehensive Public Act 26-15 in late May. Vermont enacted both an election deepfake law and a neural-data and health AI law in 2026. Several states' 2026 sessions closed with notable failures too, which matter just as much: Virginia vetoed its comprehensive AI bill, Nevada vetoed its health-insurance AI limits, and Florida's broad AI Bill of Rights passed the Senate only to die in the House. Knowing what failed is as valuable as knowing what passed, because it tells you where the political resistance sits.

#A Note on Method

Every state page is built from primary sources, the actual legislature bill pages, governors' offices, and attorney general statements, not secondary summaries. That mattered, because the secondary coverage is riddled with errors: bills reported as enacted that were actually vetoed, wrong bill numbers, failed bills cited as law. Each page is dated and will be updated as sessions move. The goal is a reference a compliance lead or a founder can actually trust.

#Why This Matters for You

If you build or deploy AI, the era of treating AI regulation as a future problem is over. It is here, it is fragmented, and the fragments contradict each other. The practical takeaways are consistent across the whole map: build disclosure into anything conversational, build consent and provenance into anything that generates a human likeness, keep a human in the loop for consequential decisions in employment, health, and insurance, and govern to the strictest state you operate in rather than the most lenient.

The same week I finished this map, the federal government switched off the most capable public AI model in the country over an export-control dispute. The lesson rhymes. Capability and compliance both now depend on a patchwork of authorities that can change with a single signature, and the only durable strategy is to see the whole board.

The full interactive breakdown, all 50 states, every tracked bill, updated as the law changes, is live now.