Why the US Doesn't Have a Federal AI Law (And What Might Change That)

In December 2024, a House task force split evenly between twelve Democrats and twelve Republicans spent ten months studying whether Congress should pass a comprehensive federal AI law. Its final report ran to 85 recommendations across national security, civil rights, and R&D. And its central procedural conclusion was that Congress should not try to write one big bill — favoring instead a narrower, sector-by-sector approach layered onto existing law. The absence of a federal AI statute isn't an accident Congress hasn't gotten around to fixing. It's the outcome of a structure that makes one comprehensive bill hard to pass, reinforced by a body that studied the question and recommended against attempting it anyway.

That's worth sitting with, because the volume of activity around federal AI policy makes the absence of a law look more like an oversight than it is. Congress has introduced close to 400 bills mentioning AI across recent sessions and held dozens of hearings. None of it has produced a floor vote on anything resembling the EU AI Act. Understanding why requires separating two different questions that get collapsed into one: why hasn't Congress passed a law, and why does the fight over federal AI policy keep resurfacing anyway, most recently as an executive order.

No committee owns "AI," so no bill does either

Federal legislation moves through committees with defined jurisdiction — Judiciary handles intellectual property and civil rights claims, Commerce handles consumer protection and communications, Financial Services handles lending and banking, Science handles research funding, and Energy and Commerce handles a chunk of health and telecom. AI touches all of it and belongs to none of it specifically. A bill regulating AI in hiring decisions is arguably Education and Workforce territory; the same bill's data-handling provisions are arguably Judiciary's; its effect on financial services underwriting pulls in a third committee entirely. There's no single "AI Committee" with the authority to move a comprehensive bill the way, say, the Judiciary Committee can move an immigration bill.

This isn't a hypothetical problem — it played out concretely. Rep. Jay Obernolte, the Republican co-chair of the House AI task force, pushed his own party's leadership to create a special committee specifically to turn the task force's 85 recommendations into actual legislation. Leadership said no. The reasoning, according to a senior GOP aide, was that the task force had already accomplished what it set out to do, and the standing committees with partial AI jurisdiction could work from the report on their own. In practice, that meant no single body was actually responsible for drafting a bill from the recommendations. The House Science Committee did advance ten AI-related bills out of committee — a real, credit-worthy step — and none of them has reached a floor vote since. That's the pattern worth understanding: these bills aren't getting voted down. They're not getting scheduled at all, which is a quieter and more durable way for legislation to die than a defeat anyone has to explain.

The real fight isn't "a federal AI law" — it's who gets to write the rules at all

Underneath the jurisdictional problem sits a second, sharper disagreement that has nothing to do with whether AI should be regulated and everything to do with who regulates it: should states keep setting their own AI rules, the way Colorado has with its AI Act, or should a single federal standard override them? That fight has now failed twice at the legislative level in 2025 alone.

The first attempt rode inside the July 2025 budget reconciliation bill — the legislative vehicle that became the One Big Beautiful Bill Act. A provision inserted by Sen. Ted Cruz would have imposed a ten-year moratorium on states enforcing their own AI laws, with a specific piece of leverage attached: states that wanted to keep enforcing AI statutes risked losing access to federal broadband deployment funding under the BEAD program. During the Senate's marathon "vote-a-rama" on July 1, 2025, an amendment from Sen. Marsha Blackburn, co-sponsored with Sen. Maria Cantwell, stripped the moratorium out entirely. The vote was 99-1. Only Sen. Thom Tillis voted to keep it in. Whatever the merits of federal preemption in the abstract, a near-unanimous Senate — across party lines — rejected doing it this way, through a budget bill, attached to broadband funding, with no separate debate on the substance.

The second attempt came later in 2025, when House GOP leadership looked at the annual National Defense Authorization Act as another vehicle for a similar moratorium. It ran into the same wall from a different direction: the chairs of the House and Senate Armed Services Committees objected to loading unrelated AI policy onto a must-pass defense bill, and a genuinely bipartisan coalition — Republican governors including Ron DeSantis and Sarah Huckabee Sanders, and senators spanning from Josh Hawley to Ed Markey and Elizabeth Warren — pushed back publicly. The preemption language didn't survive into the final FY2026 NDAA either. Two separate legislative vehicles, two different Congresses' worth of leverage, and the same outcome: Congress would not vote to override state AI law, at least not attached to something else and rushed through.

When Congress wouldn't preempt state law, the White House tried to do it by executive order instead

The preemption fight didn't end when the second legislative attempt failed — it moved to the executive branch. On December 11, 2025, President Trump signed an executive order titled "Ensuring a National Policy Framework for Artificial Intelligence," directing the Attorney General to stand up an AI Litigation Task Force within 30 days. That task force's job is to identify state AI laws the administration considers inconsistent with a "minimally burdensome" national AI policy and challenge them in court — on theories including that they impose unconstitutional burdens on interstate commerce, or that they're preempted by existing federal statutes. The order also reaches for the same lever Congress rejected twice: it conditions certain federal broadband funding on states pausing enforcement of AI statutes the administration views as conflicting with federal policy.

It's worth being precise about what this is and isn't. It is a real, live attempt to accomplish through litigation and funding pressure what two separate legislative pushes couldn't accomplish through a vote. It isn't a federal AI law. As our full history of US federal AI executive orders lays out, every executive order in this space has carried the same structural weakness: it exists only as long as the administration that issued it wants it to, and a successor can reverse it as unilaterally as it was signed — which is exactly what happened to the prior administration's comprehensive AI order in January 2025. A DOJ litigation campaign can win or lose in court on a state-by-state basis, and even a string of wins doesn't produce the kind of durable, nationwide rule a statute would. Treating this EO as the federal AI law arriving through a side door is a mistake companies building compliance programs shouldn't make.

What the uncertainty actually costs a compliance team

Consider a company like Meridian Underwriting, a fictional insurtech vendor selling AI-driven risk-scoring software to property and casualty carriers licensed in a dozen states. Meridian's general counsel has three options, and none of them is "wait for Washington to settle this." Building to a bare federal minimum doesn't work, because there isn't one yet. Building state-by-state, adjusting the product's documentation and human-oversight features for each jurisdiction's specific statute, is accurate but expensive to maintain as more states legislate. Building once to the strictest currently enacted standard — Colorado's AI Act, which imposes duty-of-care and impact-assessment obligations most other state bills don't yet match — and treating anything less demanding as already covered is the approach that scales, because it doesn't require a rebuild every time a new state bill passes or a federal fight resolves in one direction or another.

The part of this that's easy to get wrong is timing a bet on federal preemption arriving soon enough to skip the state-level work. Two legislative attempts and one executive order into this fight, over a period of less than a year, there is still no resolution — and the executive order itself is headed into litigation that could easily run past the current administration's term. A compliance program built on "the federal government will sort this out before it matters to us" has already been wrong twice in 2025 alone.

Three things that could actually produce a federal statute

None of this means a federal AI law is permanently off the table — it means the honest answer to "when" is that no one currently knows, not a prediction dressed up as certainty. Three mechanisms are genuinely plausible, though none is imminent.

The AI Litigation Task Force's court challenges could resolve the underlying question in a way that forces Congress to act regardless of who wins. A string of DOJ victories narrowing what states can enforce still leaves a gap only Congress can fill with something durable; a string of losses undercuts the executive-preemption approach and puts the pressure for a real statute back where it started, on the committees that couldn't move one before. Second, industry pressure could flip from opposing regulation generally to actively lobbying for one uniform federal standard — a distinct position from wanting no rules at all, and one that becomes more attractive to large AI developers every time a new state statute adds another compliance variant to track. Third, and least predictable, a sufficiently visible AI-caused harm could generate the kind of fast, bipartisan pressure that has historically moved federal legislation in other domains faster than years of hearings ever did. Any of these could shift the calculus. None of them has yet, which is exactly why the state-by-state and agency-by-agency landscape described in our overview of what federal AI regulation actually looks like today remains the operative reality, not an interim state waiting on Congress to catch up.