Every US State AI Law That Matters in 2026

The same AI hiring tool can be perfectly legal in Texas, a regulatory risk in Colorado, and a lawsuit magnet in Illinois. Not because the tool changed. Because three state legislatures have three completely different ideas about what "harmful AI" means. There's no federal AI law. The states aren't waiting for one. And they all disagree with each other. Colorado Is the One That Looks Like the EU full breakdown separately. Short version: if your AI makes or influences "consequential decisions" (hiring, lending, insurance, housing, education, healthcare, government services, legal services) you have obligations as either a developer or deployer. Many companies are both. Deployers need impact assessments before deployment and annually after. Consumer notifications before and after adverse decisions. A public AI transparency page. Developers have to document their systems and report algorithmic discrimination to the AG within 90 days. The penalty math: $20,000 per violation, each affected consumer counted separately. Screen 1,000 Colorado applicants with a discriminatory pattern? $20 million in theoretical exposure. But the NIST AI RMF safe harbor gives you a real affirmative defense.