I remember hitting send on a hot tip and watching the thread explode into a dozen messages. You can feel the room shift the moment federal power and local rules collide. I want to walk you through what that collision looks like and why it should make you sit up.
In a committee hearing room, a laptop sat open: What the Great American AI Act proposes
Rep. Jay Obernolte (R-CA) and Rep. Lori Trahan (D-MA) released a discussion draft this week called the Great American AI Act, described by legislators as, “intended to solicit feedback from stakeholders, experts, and the public before the bill is formally introduced.”
The draft asks large AI developers to notify the government about the development of frontier models, write plans to mitigate grave cybersecurity harms, and submit to third-party audits. It would also create the Center for AI Standards and Innovation (CAISI) inside the Commerce Department’s National Institute of Standards and Technology (NIST) and allocate $300 million (≈ €276M) to run it for the next three years.
President Donald Trump issued an executive order that aimed to set up CAISI, but without congressional funding the office has limited reach—this bill would supply the cash.
On a statehouse floor, lawmakers stared at printed statutes: How the bill would change state power
The most explosive element isn’t the audits or the test labs; it’s preemption. The draft would replace some existing and future state laws with a federal standard—a federal ceiling rather than a state floor, critics say.
That matters in places such as California and New York, where legislatures have already passed AI-related rules—often negotiated with companies like OpenAI, Meta, Google, and Anthropic. Consumer groups warn the draft would strip states of tools addressing algorithmic bias, youth mental health, deepfakes, and fraud.
What does the Great American AI Act require of AI companies?
The draft targets frontier models and asks companies to: register development timelines with federal authorities; produce incident response and security plans; subject those plans to independent audit; and allow targeted testing by CAISI. It stops short of sweeping consumer protections critics wanted.
At a kitchen table, a parent read headlines aloud: Who is celebrating—and who is furious?
Responses split predictably and not. Republicans and Democrats are listed as co-sponsors: Rep. Suhas Subramanyam (D-VA), Rep. Scott Peters (D-CA), Rep. Scott Franklin (R-FL), and Rep. Erin Houchin (R-IN) signaled early support. Yet advocacy groups blasted the draft.
Public Citizen called it a “disastrous proposal” and argued the bill ignores discrimination, youth harms, and deepfakes. J.B. Branch said the draft “does nothing” for those issues. The Tech Oversight Project released district surveys showing 56% in Obernolte’s district and 63% in Trahan’s district oppose weakening state-level protections.
Sacha Haworth of the Tech Oversight Project warned that the draft grants Big Tech “massive amnesty” from local rules. Their point: even state laws negotiated with industry were imperfect, and federal preemption could remove the only levers communities currently have.
Will this federal bill override state AI laws?
Yes—partially. The draft explicitly preempts certain state measures, trading local authority for a national testing-and-audit framework that does not yet exist in law. That trade-off is the flashpoint in the debate.
Outside a think tank panel, experts passed around printed comments: What CAISI would actually do
CAISI is pitched as a national lab to evaluate frontier models, coordinate testing, and publish standards. But critics ask: who audits the auditor? The bill would fund CAISI at $300 million (≈ €276M) and charge it with assessing risks for the next three years.
CAISI is a lighthouse whose lamp flickers—visible as a concept but fragile in a split Congress and unsure of the guardrails it will follow.
How will CAISI be funded and operate?
The proposal directs $300 million (≈ €276M) to NIST to stand up CAISI and run evaluations. The office would coordinate with agencies on national security and safety issues and create testing protocols for frontier systems, with audits and reporting requirements for companies operating those systems.
I’ve watched policy drafts grow teeth and then get neutered in committee. You should know two things: one, lawmakers framed this draft as a listening exercise; and two, the stakes are shifting from technical oversight to political leverage over state authority.
The debate collapses into a few blunt questions: do we accept a federal baseline that could let large platforms bypass local rules, or do we preserve a patchwork that lets states act faster on specific harms? The draft’s supporters say it creates a unified approach that will push U.S. leadership in AI; its opponents say it trades away immediate protections for promises of future federal tests.
If you want to file feedback, the authors asked for it: send comments to [email protected]. Lawmakers framed the draft as “intended to solicit feedback from stakeholders, experts, and the public before the bill is formally introduced.”
The bill will not live or die in a vacuum—companies and civil-society groups will flood the record, and the White House is already in the mix. I’ll be watching which voices shape the technical rules, and which shape the political dealmaking. The bill is a velvet trap for those who assumed a federal fix would be painless.
Which of those futures are you rooting for, and what would you protect first?