Federal AI Preemption Has Failed Twice. Here Comes Round Three.
A four-page policy paper vs. forty attorneys general and a 99-1 Senate vote. Place your bets.
Image created by Nano Banana 2
The White House Wants to Kill State AI Laws. Your Clients Still Have to Follow Them.
TL;DR: The White House released a four-page AI policy paper asking Congress to override most state AI laws. It has no legal force. State AI laws, including Colorado’s AI Act (effective June 30), remain fully enforceable. Your advice to clients: keep complying, build flexibility into your programs, and don’t make bets on federal preemption that has already failed twice in Congress.
You’re going to get this call. If you haven’t already.
A client, maybe a mid-size employer rolling out AI-powered screening tools, asks: “I just saw the White House wants to preempt state AI laws. Do we still need to worry about Colorado? What about the Illinois disclosure rules?”
The short answer is yes. You still need to worry about all of it.
The longer answer is where it gets interesting. And honestly, a little messy.
What the White House Actually Released
On March 20, the White House published a “National Policy Framework for Artificial Intelligence.” It’s four pages long. It lays out seven things the administration wants Congress to do. And the big one, the one that matters most to you and your clients, is federal preemption of state AI laws.
Here’s what that would mean in plain English: if Congress passes what the White House is asking for, states could no longer regulate AI model development. They couldn’t impose rules that “unduly burden” the use of AI for legal activities. And they couldn’t hold AI developers liable when someone else misuses their tools.
The carve-outs are narrow, and the fine print matters. States would keep authority over “laws of general applicability” related to child protection, fraud prevention, and consumer protection, plus zoning for data centers and their own government procurement of AI. That phrase “general applicability” is doing a lot of work. It means AI-specific state laws, like Colorado’s AI Act, might not be protected by the carve-out at all, even though they address discrimination and consumer harm. Everything else? The feds would take over.
The other six items in the paper cover ground you’d expect. Child safety protections including age verification and parental controls. A punt to the courts on whether AI training on copyrighted material violates IP law. Anti-censorship provisions aimed at government pressure on tech platforms. Regulation through existing agencies like the FTC and SEC rather than creating a new federal AI body. Regulatory sandboxes. And workforce studies. No new money for displaced workers, just studies.
But preemption is the headline. It’s the piece that will reshape (or not reshape) your compliance advice.
Why This Might Not Actually Happen
Here’s the thing nobody’s saying loudly enough: this exact idea has already failed. Twice.
Congress tried to include federal AI preemption in the GOP budget reconciliation bill, the “One Big Beautiful Bill.” Forty state attorneys general pushed back. Two hundred and sixty state legislators from all 50 states opposed it. The Senate voted 99-1 to strip it out. That’s not a close call. That’s a rout.
They tried again with the National Defense Authorization Act. Bipartisan opposition kept it out.
So this four-page paper is round three. And the political math hasn’t changed much.
The House GOP majority is razor-thin. Any preemption bill would need 60 Senate votes to clear a filibuster. And midterms are in November 2026, which means vulnerable members in both parties are going to think twice before voting to strip consumer protections their constituents actually support.
That said, preemption isn’t dead. Sen. Ted Cruz, who chairs the Senate Commerce Committee, has said the moratorium “will return.” Sen. Marsha Blackburn released her own bill draft that tracks the White House paper closely but goes further on compliance requirements. These aren’t backbenchers. They have committee power and agenda-setting authority.
On the other side, House Democrats introduced a bill the same day the paper dropped to repeal the president’s December executive order on AI, with a Senate companion bill planned. So we’ve got active legislative fights on both flanks.
The honest assessment: federal preemption is possible but not probable in this Congress. And even if something passes, legal challenges would follow immediately.
The State Laws That Are Actually Enforceable Right Now
While Washington debates what might happen, state laws are already happening. And they’re coming fast.
Colorado’s AI Act is the big one. It takes effect June 30, 2026. That’s roughly three months from now. It requires “reasonable care” to protect against algorithmic discrimination in high-risk AI systems. And the list of high-risk areas reads like a client intake form: employment, housing, credit, education, healthcare, insurance, legal services. The White House has specifically targeted this law, and the DOJ’s AI Litigation Task Force (created by the December 2025 executive order) was designed in part to challenge it. But as of today, no injunctions have been obtained.
California’s AB 2013 has been live since January 1. It requires AI developers to disclose the data used to train their systems, and it applies broadly enough that your clients’ vendors will feel it even if your clients don’t directly. Illinois amended its Human Rights Act to require employer disclosure when AI is used in hiring and to prohibit discriminatory AI-driven employment decisions. Texas passed its own AI governance act requiring consumer-facing disclosure and banning intentional discrimination through AI systems. And New York City’s Local Law 144 continues to require bias audits for automated employment decision tools.
Every one of these is enforceable today (or will be by summer). None of them are suspended by a White House policy paper.
The Real Compliance Challenge: Uncertainty Itself
Here’s what I keep coming back to. The hardest part of this for attorneys isn’t understanding the policy document or tracking the state laws. It’s advising clients through genuine uncertainty.
Your clients can’t retire their state compliance programs because a four-page document says the administration wants Congress to do something. That’s not how law works. But they also can’t pretend the ground isn’t shifting. If preemption does pass, companies that over-invested in state-by-state compliance architectures might have wasted significant money. If it doesn’t pass, companies that pulled back on compliance are exposed.
This is a “comply now but stay flexible” moment. Which, I know, is the least satisfying advice to give. But it’s the honest one.
What to Tell Your Clients This Week
State AI laws are still the law. Colorado, California, Illinois, Texas, NYC, and others remain fully enforceable. A White House policy paper is not legislation. Don’t let clients confuse the headline with a rule change.
Build compliance programs that can bend. Design AI governance policies in modular pieces, not monolithic state-by-state silos. If preemption passes later, you want to adjust without starting over. If it doesn’t, you’re already covered.
Get ready for June 30. Colorado’s AI Act effective date is the next real deadline. If your clients use AI in any of the high-risk categories (employment, housing, credit, education, healthcare, insurance, legal services), they need a plan before that date. Not after.
The Bottom Line
We’re in a gap period. The federal government wants to preempt state AI laws but hasn’t been able to get Congress to agree. The states have already acted. And attorneys are stuck advising clients through the space between those two realities.
That gap won’t close soon. But the Colorado deadline will.
Get your clients ready for June 30. That’s the thing you can actually control.
If you read this far, you’re probably the person at your firm who gets the call when a client forwards a headline and asks, “Does this change anything?” And right now, the honest answer is harder to give than the easy one.
That’s the conversation I have every day with firm leaders who are trying to build compliance programs on ground that keeps shifting. If you’re working through what Colorado means for your clients, or how to design AI governance that doesn’t have to be rebuilt every time Washington releases a new paper, I’d like to hear what you’re sitting with. Reach me at steve@intelligencebyintent.com. I’ll tell you what I’m seeing, what’s working, and what isn’t ready yet.