Rulebook with a floor
National rulebook with a safety floor. One federal AI law replaces state laws, but keeps safety plans, reports, incident reporting and audits.
What it does
Congress replaces state frontier-AI laws, like California's SB 53 and New York's RAISE Act, with one federal law. The federal law keeps a floor: published safety plans, transparency reports, incident reporting and outside audits for the largest developers.
Applies to: The largest frontier AI developers, nationwide.
- Replace state laws on frontier AI development with one federal law.
- Require the largest developers to publish safety plans and transparency reports.
- Require incident reporting and outside audits under federal enforcement.
Where things stand
Frontier-AI safety rules today are state laws, California's SB 53 and New York's RAISE Act, plus an executive order pressuring states not to regulate. No federal law sets either preemption or a safety floor.
Why it costs almost no lead
Most of what it requires, the big labs already do under state law and their own safety plans. One rulebook instead of several saves a little compliance work: about a day gained.
Biggest unknown: How strong the federal floor ends up, and whether it keeps pace as state laws would have.
Why it lowers p(doom) by ~0.03%
Keeps a safety floor of plans, reports and audits in place nationwide while replacing state laws with one federal standard.
It trades state laws for a national floor. Nationwide coverage and audits add a little; losing stricter state rules and state experiments takes some away. On balance a slight improvement, unlike preemption with no floor, which raises risk.
The strongest case that it costs more
Preemption removes state protections now, while federal enforcement takes years to build. A single national floor is easier for industry to capture and freeze at a weak level than many state experiments, and SB 53 and RAISE already provide most of what it offers.
The debate
For
- Dean Ball proposed that Congress preempt state laws aimed at AI developers while requiring federal disclosure of frontier safety practices, 2025.
- Anthropic said Congress should replace state laws only with a federal law at least as strong as its proposed transparency framework, 2026.
- Reps. Jay Obernolte and Lori Trahan introduced the FRONTIER Act, with transparency, safety frameworks, audits, incident reporting and preemption of state frontier-safety laws, 2026.
- Foundation for American Innovation called the FRONTIER Act Congress's best AI bill yet, praising its narrow preemption, 2026.
Against
- 36 state attorneys general urged Congress to reject federal preemption of state AI laws, saying it would undermine states' ability to respond to AI harms, 2025.
- Americans for Responsible Innovation argued that the FRONTIER Act's preemption takes effect before federal oversight exists and gives whistleblowers no channel to report violations, 2026.
- The White House sought broad preemption of state AI laws under a 'minimally burdensome' standard with no frontier-safety requirements, the opposite side of this debate, 2026.
Sources
Rough starting points, not precise forecasts. Lead costs assume China doesn't depend on U.S. models, the case least favorable to safety laws, and count 3 years. On the menu you can change every assumption and put in your own numbers. Last priced 2026-09-26.