US State AI Laws · Overview

US state AI laws: an overview

Hael · Updated 6 July 2026 · 6 min read

Key takeaway

In the absence of a single comprehensive federal AI law, individual US states have begun setting their own rules for artificial intelligence. The result is a growing patchwork: different states with different requirements, effective dates, and definitions. For any organisation that builds or deploys AI across the United States, this means watching several state laws at once rather than a single national standard.

With no broad federal AI law, US states are setting their own AI rules, creating a patchwork.
The key states to watch include Colorado, Texas, California, and New York City.
A single AI system can be subject to several state laws at once, with differing requirements.
A coherent governance foundation adapts to multiple state laws far more easily than separate efforts.
Current as of July 2026. The Colorado position is in active motion; this guide tracks the live state.

Why a patchwork exists

The United States has not enacted a broad federal AI law comparable to the EU AI Act. Federal activity has centred on executive action, agency guidance, and sector rules rather than a single statute governing AI across the economy. Into that gap, states have stepped, much as they did with data privacy after California's privacy law. The consequence is that AI regulation in the US is, for now, defined substantially at the state level.

The key states to watch

Several states have moved furthest, and they are the ones most organisations need to understand:

Colorado: repealed its original broad AI law (SB 24-205) in May 2026 before it took effect, replacing it with the narrower ADMT Act (SB 26-189), a disclosure law effective January 2027.
Texas: Enacted the Texas Responsible Artificial Intelligence Governance Act (TRAIGA), addressing certain AI uses and harms.
California: Active on multiple fronts, including rules on automated decision-making and AI transparency.
New York City: Local Law 144 governs the use of automated employment decision tools, an early and influential measure.

Many other states have introduced or are considering AI legislation, so the landscape continues to expand.

What this means in practice

For organisations operating nationally, the practical challenge is coverage. A single AI system might be subject to Colorado's rules for some users, California's for others, and New York City's if used in hiring there. The obligations differ, but the underlying disciplines, knowing your AI systems, classifying them, governing high-risk uses, and being transparent, are common across them. An organisation with a coherent AI governance practice can adapt it to multiple state laws far more easily than one starting fresh for each.

How to approach the patchwork

The efficient approach is not to treat each state law as a separate project but to build a governance foundation that covers the common ground, then map the specific requirements of each relevant state onto it. Because the state laws, the EU AI Act, and frameworks like the NIST AI RMF share a great deal of substance, governing your AI well once positions you to meet many of them. The patchwork is easier to manage as a single, adaptable practice than as a stack of separate compliance efforts.

Key terms

Patchwork: A landscape of overlapping state laws in the absence of a single federal one.
Federal AI law: A national US statute governing AI; not yet enacted as a broad framework.
Covered ADMT: Automated decision-making technology processing personal data to guide consequential decisions, the focus of Colorado's SB 26-189.
AEDT: Automated employment decision tool, regulated by NYC Local Law 144.

References

Related guides