State AI Laws Are Multiplying. That's Actually Evidence Against Federal Legislation Passing.
TexTak holds [us-federal-ai-law] at 18% — one of our lower-confidence forecasts — on the thesis that partisan gridlock and the Trump administration's deregulatory posture make comprehensive federal AI legislation unlikely before midterms. Today's news from California, Vermont, Alabama, and Hawaii complicates the picture in both directions: the state-level surge is real evidence of regulatory momentum, but it may actually be working against our bull case for federal legislation rather than supporting it. This is the piece where we genuinely wrestle with whether we have the causal arrow right.
The traditional argument for federal preemption legislation goes like this: patchwork state laws create compliance chaos for companies operating nationally, companies lobby for a single federal standard, Congress responds. Vermont's synthetic media disclosure law, Alabama's health insurance AI regulation, Hawaii's chatbot safety bills, and California's aggressive committee pipeline are all consistent with the first step of that chain. The patchwork is real and accelerating. Cruz and Obernolte are actively drafting preemption legislation. The logic seems to be building.
Here's what keeps us anchored at 18% despite that logic: the compliance chaos argument for federal preemption has existed for tech regulation for over a decade — for privacy, for platform liability, for data security — and it has never produced a comprehensive federal law. Companies adapt to state-by-state compliance. The lobbying effort required to pass federal legislation is not obviously less costly than the compliance effort required to navigate state laws. More importantly, the Trump administration's deregulatory posture creates a structural problem: any federal AI bill that passes would need to either preempt state laws (which requires strong federal standards, which the administration opposes) or create a floor that states can exceed (which doesn't solve the patchwork problem). Neither configuration has a clear congressional majority.
The genuine complication in today's evidence is that California's legislative push is aggressive enough that it might force the issue. California has functionally set national standards before — automotive emissions, privacy under CCPA — and if California's AI bills pass in their current form, companies face a de facto national standard regardless of what Congress does. That actually reduces the urgency for federal action, not increases it. If California does the work, Congress doesn't have to.
We're holding at 18%, but we're watching one specific variable: whether Nvidia, Google, Meta, or Microsoft shift from passive preference for federal preemption to active, coordinated lobbying with specific bill language. That shift — not more state bills passing — is what would move us toward 30%. The state activity is proximate evidence that conditions for federal legislation exist. It is not direct evidence that federal legislation will pass. Those are different things, and we want to be clear about which we're observing.