AI Regulation Tracker  /  Regulation in force

Four States, Four AI Hiring Rulebooks: How to Build One Workflow That Survives All of Them

Illinois, Texas, Colorado, and now Connecticut each regulate AI in hiring differently. A national employer cannot run four processes. Here is the single documentation workflow that builds to the strictest standard of all four.

By Anthony Guerriero  ·  Reviewed by The Leveraged Years Editorial Desk  ·  Published June 28, 2026  ·  Last updated June 28, 2026

As of 2026, Illinois, Texas, Colorado, and Connecticut each regulate AI in hiring under different rules (Illinois HB 3773 and Texas HB 149 effective January 1, 2026; Connecticut SB 5 operative October 1, 2027; Colorado SB 26-189 effective January 1, 2027). A multi-state employer does not need four processes: build one documentation packet to the strictest standard (disclosure, pre-decision notice, vendor-oversight file, bias-testing record) and apply it everywhere. Primary sources: the four state statutes.

The patchwork is now a real operational problem

For two years, multi-state AI hiring law was a forecast. As of 2026 it is a desk problem. Illinois HB 3773 took effect January 1, 2026. Texas TRAIGA took effect the same day. Connecticut signed SB 5 on May 27, 2026, with its automated employment provisions operative October 1, 2027. Colorado scrapped its original AI Act and replaced it with SB 26-189, a disclosure-based regime effective January 1, 2027. Four states, four rulebooks, four sets of definitions, and one employer that hires across all of them.

The instinct to run a separate process in each state is the wrong one. It is expensive, it is fragile, and it guarantees that the day a recruiter in one state follows the wrong checklist is the day a violation happens. The better move is to find what the laws share and build to that.

What the four laws actually have in common

Strip away the differences and a shared pattern appears. These regimes lean on four recurring ideas.

First, notice that a tool is in use. When an automated system interacts with or evaluates a candidate, the laws converge on telling the person that it is happening. Connecticut requires plain-language disclosure when an automated tool interacts directly with an applicant. Illinois requires notice when AI is used in hiring decisions.

Second, pre-decision transparency. Before a consequential employment decision, the stricter laws want the affected person to know the tool was used and what it considered. Connecticut spells this out in detail: the trade name of the tool, the categories of personal data it weighs, and where that data comes from.

Third, recordkeeping. Each regime assumes the employer can produce documentation of the tool, its purpose, and its handling of personal data. You cannot disclose what you have not recorded.

Fourth, accountability for bias. Connecticut amended its discrimination statute so that using a tool is not a defense to a bias claim, while allowing evidence of anti-bias testing to help. New York City's Local Law 144 already requires bias audits. The direction is unmistakable: document that you checked the tool for disparate impact, because the absence of that record is its own exposure.

Build to the strictest standard, then apply it everywhere

Here is a workflow built to the strictest standard among the states you operate in and built to meet or exceed what each law asks. This is a practical synthesis, not statutory harmonization, so state-specific counsel review remains essential. Build one packet and apply it everywhere.

The packet has four documents. A candidate-facing disclosure that states an automated tool may be used in the process and what kind of decision it informs. A pre-decision notice template that names the tool, its purpose, the categories of data it analyzes, and the data sources, ready to issue before an adverse decision. A vendor-oversight file for each tool, capturing who vetted it, what data it touches, and the documentation the developer is contractually required to provide. And a bias-testing record showing the tool was assessed for disparate impact, with dates and results.

Running the strictest version everywhere costs a little more in the lightest-regulation states and saves enormously in avoided complexity. One process is auditable. Four are not.

Why national employers should not wait for Connecticut's 2027 date

Connecticut's operative date is October 1, 2027, which tempts a wait-and-see posture. That is a trap. Illinois and Texas are already live, the documentation the packet requires takes months to assemble across HR, legal, and vendors, and the trend line is adding states, not subtracting them. The employer that builds the packet now is ready for the next state law before it is signed, because the next law will very likely ask for some subset of the same four things.

The patchwork is not going to resolve into a single federal rule soon. The way to beat it is not to track every statute. It is to build one defensible process that already contains what they share.

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.