Colorado Gutted Its AI (Hiring) Law. Washington Didn’t.

Happy Thursday, Job Board Doctor friends!

Two stories this week that, read together, tell you something uncomfortable about where worker protections actually come from. Both are happening at the state level, because the federal level is not happening at all, or rather, the federal level is happening, but against the states rather than for the workers.

They are moving in opposite directions, and the reason why is the whole story.

• Colorado built the country’s toughest AI hiring law, then took it apart.
• Washington quietly expanded Fair Chance protections that will take effect July 1.
• What both say about regulation by patchwork, and where the federal government stands.

Let’s dive in.

Colorado: The Law That Got Gutted

In 2024, Colorado passed SB 24-205, the first comprehensive state AI law in the country.  The law required developers and deployers of high-risk AI systems, explicitly including systems used in employment decisions, to use reasonable care to protect people from algorithmic discrimination.

Reasonable care, as defined by the Legal Information Institute at Cornell University Law School, is, “The level of care that a reasonable person would exercise in such circumstances. Failure to exercise reasonable care may lead to liability, if such a failure caused an injury; while exercise of reasonable care can establish that a party acted reasonably and is not liable”.

It borrowed its architecture directly from the EU AI Act: a risk-based regime built on a duty of care, risk management programs, and impact assessments (Seyfarth).

For anyone selling or buying AI hiring tools this legislation was one to watch. 

The law was originally set to go live February 1, 2026. A special session in August 2025 pushed that to June 30, 2026, after lawmakers couldn’t agree on amendments. Then, per Norton Rose Fulbright, xAI sued to block the law in early April, the federal government intervened, and on April 27 the Colorado Attorney General agreed to suspend enforcement.

Two weeks later, the legislature passed SB 26-189, which Governor Polis signed May 14. The new law repeals and replaces the original.

According to David Stauss, writing for Troutman, Pepper Locke, it strips out the duty of care, the risk management programs, and the impact assessments, replacing all of it with a disclosure-and-transparency framework focused on automated decision-making technology. The new law now takes effect January 1, 2027.

There is a federal hand worth naming here, even if it did not do the killing. On December 11, 2025, the White House issued an executive order, Ensuring a National Policy Framework for Artificial Intelligence, directing the Attorney General to stand up an AI Litigation Task Force whose sole job is to challenge state AI laws the administration considers inconsistent with its policy, on theories including unconstitutional burdens on interstate commerce and federal preemption.

Economic Policy Institute tell us the order names Colorado’s algorithmic-discrimination law by reference. The task force began operating in January 2026. By April, per Seyfarth, the Department of Justice had intervened in xAI‘s suit against the Colorado law, and the Attorney General agreed to the enforcement stay shortly after.

Two caveats – as I always want to hold myself to account.

The executive order preempts nothing on its own; legal analysts across the board are clear that only Congress or the courts can override a state law, and the governors of California, Colorado, and New York said publicly the order would not stop them.

The Colorado repeal was already in motion for its own reasons. So the accurate read is not that Washington, D.C., dismantled Colorado’s law. It is that an corporate aligned repeal already underway got a federal assist.

One item worth noting – This was not a party-line fight.

It was a working bipartisan majority concluding that the EU-style model was the wrong vehicle, under heavy pressure from the tech industry, an active lawsuit, and a federal government that had built permanent machinery to make exactly this kind of pressure routine.

The outcome here is pretty straightforward. The most aggressive attempt by any state to regulate AI, including in hiring, did not fail because it was poorly drafted or because protections weren’t needed.

It was unwound because a single state legislature, is exactly the structure that industry, litigants, and now a dedicated federal task force can converge on and bully into subjugation. 

Washington: Protections That Actually Landed

Now contrast that with what happened in Washington, on a different kind of protection that draws far less venture capital attention and no federal task force at all.

Washington already had a Fair Chance Act limiting how employers use criminal history. In 2025 it expanded it substantially through HB 1747. Employers cannot run a criminal background check until after making a conditional offer. (Fisher Phillips)

That is a real change from the prior standard, which let employers inquire once a candidate was deemed otherwise qualified.

The amended law also bars blanket exclusions of anyone with a record, prohibits adverse action based on arrest records or juvenile convictions, and requires a documented legitimate business reason for rejecting someone over an adult conviction.

And it has teeth: penalties run from $1,500 up to $15,000 per violation, enforced by the Attorney General.

Washington’s Fair Chance expansion is, in headcount terms, a far more sweeping intervention into hiring than Colorado’s AI law would have been (at least for now).

It touches every employer in the state and constrains a screening step nearly all of them use. But it advanced on schedule, with teeth intact, and no one filed a federal lawsuit to stop it.

The difference is not the size of the protection. The difference is that fair chance hiring is not where the AI industry has staked its lobbying budget, and background-check timing does not trigger a compelled-speech argument or a federal task force the way algorithmic-discrimination duties did.

Protections advance where the most heavily capitalized opposition, and now the federal government, isn’t standing in the way.

The EU Comparison and the Patchwork Problem

The instinct is to hold up the EU AI Act as the adult in the room and it still is despite some delays pushing implementation forward.

On May 7, 2026, the EU reached a provisional agreement to delay its high-risk obligations, the ones covering employment AI used for recruitment, screening, promotion, and termination, from August 2, 2026 to December 2, 2027.

According to Gibson Dunn, the 16-month slip, is driven in part by technical standards arriving late and most member states missing the deadline to even designate their enforcement authorities.

The distinction worth holding onto: the EU delayed its rules, it did not repeal them.

Employment AI is still high-risk under Annex III, and the substantive obligations (transparency, human oversight, data governance, logging) survive intact. Colorado, by contrast, repealed the duty of care outright.

That last point is the one for this audience.

There is no federal AI hiring law and no federal fair-chance standard. What workers and employers actually get is determined entirely by which state they are standing in, and increasingly by whether the federal government decides to intervene against a given state’s attempt.

For anyone running a job board, an ATS, or any tool that touches a hiring decision, this is not an abstraction. A compliance obligation that exists in Washington does not exist in Texas. An AI-screening duty that was about to bind in Colorado evaporated in six weeks. The product has to be built for fifty regimes that do not agree and keep changing, and the candidate’s protections depend on geography rather than principle.

The pattern across all three jurisdictions is the same mechanism producing different outcomes. Where regulation is low-profile and the opposition diffuse, as with Washington’s background-check rules, protections advance and stick.

Where regulation is high-profile and the opposition concentrated, well-funded, and now federally backed, as with AI hiring rules in both Colorado and Brussels, the timeline slips or the law gets hollowed out. The protection that survives is not necessarily the one workers need most.

It is the one nobody with a lobbying budget, or a litigation task force, bothered to fight.

That is the week. One state took apart the toughest AI hiring law in the country under industry and federal pressure, another quietly expanded protections that will reshape how every employer in it screens candidates, and the federal government continued to offer workers nothing while building standing infrastructure to challenge the states doing the work. Build accordingly.

As always, tell me what you think, what you know, and anything I got wrong. We learn together.

Until Next Time,

Julie “The Doc” Sowash

[Want to get Job Board Doctor posts via email? Subscribe here.]

[Got a tip, document or intel you want to share with the Doc? Tell me. Tip so hot you need it to be encrypted? Use Signal.]