While Washington Deliberates, the States Wrote America’s AI Rulebook: 280 AI Laws in 2026

If you want to know what AI regulation looks like in America, do not watch Washington. Watch Sacramento, Albany, Denver, Springfield and Austin. According to fresh data from the National Conference of State Legislatures, US state lawmakers introduced 1,539 AI bills this year and enacted or adopted 280 of them through early October. A year ago the numbers were 1,262 introduced and 227 enacted. In a country with no comprehensive federal AI law, the rulebook is being written fifty states at a time.

That pattern is not new, but 2026 accelerated it. Congress has talked about sweeping AI legislation for years. The states, meanwhile, just pass things. The result is a patchwork that anyone building or selling AI in the US has to take seriously, because state attorneys general enforce these laws with real enthusiasm.

The big shift: telling people they are talking to AI

The most striking change in 2026, says NCSL Fellow Will Clark, is in notification bills. Proposals requiring companies to disclose when a person is interacting with AI nearly tripled this year. The idea sounds modest: a chatbot that says it is a bot, a synthetic voice that identifies itself. But for companies running AI customer service agents, hiring tools or marketing automation across state lines, it means auditing every touchpoint where a model talks to a human.

Deepfake legislation was the most common type of AI law actually enacted in 2026, typically covering explicit content, political advertising and the unauthorized use of someone’s digital likeness. Other enacted measures targeted AI in health care and education. The pattern is consistent with 2025, when private-sector AI use and deepfakes topped the agenda, but the volume and specificity keep growing.

Only two states went comprehensive

For all that activity, only two comprehensive AI bills became law in 2026: Colorado’s SB 189 and Connecticut’s SB 5. Colorado has been the leader since its 2024 AI Act on high-risk systems, and SB 189 extends that framework. The rest of the states mostly chose targeted laws: one problem, one statute. That keeps bills passable, but it also means companies face dozens of overlapping definitions of harm, disclosure and consent.

Legislation aimed at the biggest players was more common than in past years. Illinois passed SB 315, which mandates third-party audits for frontier models, and New York’s SB 8828 sets safety and transparency requirements for large AI developers. The message to labs is clear: if you train the models, you get the obligations, and the scrutiny comes from Albany and Springfield rather than Washington.

What Washington is actually doing

The federal government has been less active than the states, Clark says. Congress has deliberated several bills, including the Kids Online Safety Act, the App Store Accountability Act and the Great American AI Act. The last of the three would require large AI developers to create, implement and publish plans addressing potential catastrophic risks from their models.

But federal activity is not just slower, it is aimed in the opposite direction. A bipartisan working draft of the Great American AI Act from House lawmakers would preempt certain state AI laws for three years. NCSL has said it is deeply concerned about any federal legislation that would preempt states’ ability to protect their residents. A coalition of 26 state attorneys general is pushing Congress for federal AI safety rules that explicitly do not preempt state law and that give state enforcers authority under any future statute.

So the coming fight is not only about how strict AI rules should be. It is about who gets to set them at all. States that spent two years building AI enforcement muscle are not volunteering to stand down.

What this means if you build AI products

The practical consequences are already here. If you sell an AI product in multiple states, you are navigating disclosure rules in one jurisdiction, chatbot safety standards in another, deepfake liability in a third, and audit obligations for frontier models in Illinois and New York. The narrowest laws are the easiest to miss: a notification requirement sounds trivial until an attorney general turns it into an enforcement sweep.

Compliance teams are treating the NCSL Artificial Intelligence Legislation Database, updated monthly, as required reading. The smartest companies are building disclosure and documentation into their products by default, because retrofitting a chatbot to announce itself in one state but not another is the kind of complexity that breaks teams.

There is also a strategy lesson in the numbers. When 280 AI measures become law in a single year without any federal framework, lobbying only in Washington misses where the action is. The AI industry spent heavily to shape federal policy. The states just outlegislated them.

The race is not slowing down

With midterm elections on November 3, AI safety is now a campaign issue, and a fresh legislature in 2027 will almost certainly introduce more bills than this year’s 1,539. The White House has leaned on voluntary agreements with the big labs, but six states have sued OpenAI this year, including a coordinated October suit by eight attorneys general. Voluntary does not look like where this ends.

The American AI rulebook of the next decade is being drafted right now, in statehouses most AI founders have never visited. The companies that pay attention to those fifty capitals will be the ones that survive the patchwork. The ones that don’t will find out the hard way that the most important AI regulator in their business is a state attorney general they have never heard of.

Frequently Asked Questions

How many AI laws have US states enacted in 2026?

US state lawmakers introduced 1,539 AI bills and enacted or adopted 280 through early October 2026, according to the National Conference of State Legislatures. That compares with 1,262 introduced and 227 enacted in 2025.

Which US states have comprehensive AI laws?

Colorado and Connecticut enacted the only two comprehensive AI bills in 2026: Colorado SB 189 and Connecticut SB 5. Illinois (SB 315) and New York (SB 8828) targeted frontier models specifically, with mandated third-party audits and safety and transparency requirements.

Is there a federal AI law in the United States?

No comprehensive federal AI law exists. Congress is considering the Kids Online Safety Act, the App Store Accountability Act and the Great American AI Act. A working draft of the latter would preempt certain state AI laws for three years, a provision the National Conference of State Legislatures strongly opposes.