Political Tech
Washington wants to own AI regulation. State capitals aren't waiting for permission.
NewsOnScale Staff
July 1, 2026
There is a familiar argument being made in Washington right now: that artificial intelligence is too important, too technically complex, and too nationally consequential to be left to fifty different state legislatures. The federal government, the argument goes, needs to set the floor — and ideally the ceiling — for how AI systems are built, deployed, and held accountable.
States are not buying it.
Over the past eighteen months, more than two dozen states have introduced or passed legislation touching AI in some form — covering everything from algorithmic hiring tools to deepfake disclosure requirements to automated decision systems used in housing and credit. Colorado, Texas, and Illinois have moved particularly aggressively. California, despite the high-profile failure of SB 1047 last year, continues to generate new proposals. The legislative momentum at the state level is not slowing down.
This is the central tension that has emerged as the defining structural fight in AI governance: not whether to regulate AI, but who owns that authority.
## What Centralization Actually Means
The push from the White House and certain federal agency coalitions toward consolidated AI oversight is often framed as a matter of coherence and competitiveness. Fragmented state laws, proponents argue, create compliance nightmares for companies operating nationally, and could put American AI developers at a disadvantage relative to international competitors operating under unified national frameworks.
That argument has real merit in narrow contexts — there are genuine costs to building fifty different compliance programs for the same underlying technology. But coherence and federal control are not the same thing, and conflating them papers over a harder question: coherent toward what values, and accountable to whom?
Federal AI governance, as currently envisioned, would primarily flow through agencies like the FTC, NIST, and sector-specific regulators. These bodies operate with significant insulation from direct democratic accountability. State legislatures, whatever their limitations, are closer to the voters who experience the consequences of AI deployment in their workplaces, courts, schools, and housing markets.
## The Preemption Risk
The sharpest edge of this debate is federal preemption — the legal mechanism by which federal law can override state action. Depending on how federal AI legislation is eventually structured, it could effectively nullify stronger state-level protections, locking in a regulatory floor that becomes, in practice, a ceiling.
This is not a hypothetical concern. It is precisely what happened with certain provisions in federal financial regulation, where preemption of state consumer protection laws contributed to conditions that enabled the 2008 mortgage crisis. Critics of federal AI preemption are not being alarmist when they raise this history — they are citing a documented pattern.
If a federal framework sets baseline transparency requirements for algorithmic decision systems, for example, but preempts states from requiring stronger audit rights or private rights of action, the practical effect may be to weaken accountability, not strengthen it.
## Who Bears the Cost of Getting This Wrong
The communities most likely to experience the harms of ungoverned AI — workers subject to algorithmic management, tenants screened by automated systems, defendants assessed by pretrial risk tools — are also the communities with the least lobbying presence in federal rulemaking processes.
State legislatures, however imperfect, have been more responsive to this constituency in the current cycle. That is partly because the harms are more visible at the local level, and partly because state attorneys general and civil rights organizations have been more aggressive in surfacing them.
A federal framework that does not explicitly preserve meaningful state authority to go further is not a governance solution — it is a liability shield dressed in the language of coordination.
## What to Watch
The next six months will be clarifying. Congressional appetite for standalone AI legislation remains uncertain, but agency-level rulemaking is accelerating. The specific preemption language in any federal proposal — whether explicit, implied, or absent — will tell you more about whose interests are being served than any preamble about innovation or safety.
State legislatures will continue moving regardless. The question is whether federal action empowers that work or forecloses it.