The Federal Clout: US House Lawmakers Release Draft Bill to Prohibit State AI Rules

The fight over who gets to govern AI in the United States has landed squarely in Congress. On June 4, 2026, Reps. Jay Obernolte (R-Calif.) and Lori Trahan (D-Mass.) released a 269-page bipartisan discussion draft  the Great American Artificial Intelligence Act of 2026. At its center sits a high-stakes trade: a three-year federal freeze on state laws regulating AI development, in exchange for mandatory national safety plans, audits and penalties for the largest frontier labs. But the claim circulating that states would lose all AI authority is wrong  regulation of how AI is used stays with the states. Here is what’s real, what’s exaggerated, and what happens next.

At a Glance

• Draft: Great American Artificial Intelligence Act of 2026  a 269-page discussion draft released June 4, 2026 (Politico, Axios)
• Core proposal: 3-year federal preemption of state laws governing AI model development (sunset clause; renewable only by Congress)
• What states keep: authority over AI use/deployment  employment, privacy, consumer protection and healthcare laws stay intact
• Who’s covered: “large frontier developers” with more than $500 million in gross revenue in the prior calendar year
• New institution: formally establishes CAISI inside the Commerce Department; $300 million over three years
• Opposition: safety advocates call it a “generational mistake” that turns protections’ floor into a ceiling
• Status: still a discussion draft  not a formally introduced bill; analysts see little chance of movement before the August recess

Why This Matters

US AI regulation currently sits in an awkward middle: California, New York and Illinois have each passed frontier-safety laws, while companies argue that 50 different state regimes would be impossibly expensive to follow. Safety experts counter that states have been the first line of defense for tech accountability. This draft is the first serious attempt to bind both sides into one framework  and if it becomes law, it redraws the entire map of who sets AI rules in America. One question decides everything: will the federal standard be a floor, or a ceiling?

Main Story

1. What the draft actually does. Released on June 4, the discussion draft rests on two pillars. First, it preempts state laws that “specifically regulat[e] the development” of AI models for three years  during which California, New York and Illinois frontier-safety statutes would effectively be “federalized” (Roll Call, Broadband Breakfast). Second, it creates a mandatory national safety architecture: large developers must publish public “frontier AI frameworks,” report critical safety incidents, and submit to third-party auditors.

2. Who it covers. The bill is not aimed at ordinary businesses. It targets “large frontier developers”  AI companies with more than $500 million in gross revenue in the previous calendar year. If their models could pose a “catastrophic risk”  defined as a foreseeable risk of death or serious injury to more than 50 people, or over $1 billion in property damage, via weapons of mass destruction, cyberattacks, or harmful autonomous action without meaningful human oversight  they must publish how they identify and manage those risks. Violations carry penalties of up to $1 million per day.

3. Who enforces it. The draft formally establishes the Center for AI Standards and Innovation (CAISI) within the Commerce Department’s National Institute of Standards and Technology, funding it at $300 million over three years ($100 million annually, 2027–2029). Notably, OpenAI, Anthropic and Google DeepMind already partner with CAISI on voluntary model evaluations. The bill also includes whistleblower protections for AI employees, tougher criminal penalties for AI-assisted fraud, workforce research and education programs.

📋 Table 1: What Changes If the Draft Becomes LawStatus Quo vs Proposed Framework
Area	📜 Current situation	🚀 If the draft passes	Why it matters
📋 Regulatory map	Patchwork of state laws (California, New York, Illinois and more)	✅ One national standard for AI development (3 years)	AI companies operating across states
🛡️ Safety planning	Voluntary company commitments	✅ Public safety plans mandatory for large developers	Investors and regulators
👁️ Oversight	State agencies / attorneys general	✅ CAISI + third-party auditors	Frontier labs
💰 Penalties	Vary by state	⚠️ Up to $1M per day for safety violations	Developers with $500M+ revenue
🏛️ States' role	Can legislate development and use	Development frozen 3 years; use authority intact	State governments and consumer groups

(Sources: Roll Call, Axios, Broadband Breakfast; the draft is still at discussion stage and final text may change)

4. Supporters vs critics  where the fight really is. In an op-ed, Obernolte and Trahan argued that rather than leaving protections “only in a handful of states or forcing innovators to navigate dozens of different legal regimes,” the framework would “establish one national standard.” Speaker Mike Johnson backs preemption; industry-adjacent NetChoice praised it as a “strong” bipartisan framework. But the pushback is fierce. Brad Carson, president of Americans for Responsible Innovation, called preemption “a generational mistake” that turns “the current floor on state AI legislation” into “a federal ceiling.” Brendan Steinhauser of the Alliance for Secure AI said a national standard should “protect at least as much as it preempts.” State lawmakers in Massachusetts and New York warned Trahan against preempting their authority; Colorado’s attorney general has previously threatened to sue over federal preemption moves, and attorneys general from 22 states and DC have pushed back on similar efforts.

📋 Table 2: What States Lose  and What They Keep  Under the Draft
Area	Under the draft	Explanation
🏛️ Laws on AI model development	❌ 3-year federal preemption	States can't pass new development rules; authority returns after sunset unless Congress renews
⚖️ Laws on AI use/deployment	✅ Stays with states	Employment, privacy, consumer protection, healthcare, financial-services rules untouched
🚀 Frontier safety laws (CA, NY, IL)	⚠️ Would be "federalized"	Existing state frontier laws fold into the national framework
🛡️ Enforcement	🟡 State AGs retain enforcement role	The draft document says state attorneys general would retain authority to enforce the federal standards

5. Claims vs reality. Not everything circulating around this announcement holds up. The idea that states would be “completely prohibited” from AI rules misreads the draft: preemption lasts three years and covers development rules only. The rollout mentions of “California, New York and Texas” don’t match the bill either  the text names California, New York and Illinois. And it is not law: it is a discussion draft soliciting feedback. Legal analysts note most ordinary businesses would see little direct impact, since the obligations target the biggest frontier developers.

📋 Table 3: Claims vs Verified Facts
Circulating claim	Status	Verified fact
"States completely and permanently banned from AI rules"	❌ Misleading	Preemption covers development rules only, for 3 years (sunset clause); deployment authority stays with states
"California, New York, Texas targeted"	⚠️ Partially wrong	Bill names California, New York and Illinois; Texas is not cited
"A single supreme federal framework"	🟡 Partly true	One standard for development; deployment rules, existing executive orders and NIST guidance remain
"Mandatory risk assessments for high-risk foundation models"	✅ True	$500M+ developers must publish frontier AI frameworks and report critical incidents
"The draft is already law"	❌ False	Discussion draft only; no formal introduction yet
"Regulatory clarity for tech and VC"	🟡 Analysis	One standard should cut compliance costs, but final text and passage remain uncertain

6. What happens next. Politico frames this as Republicans’ last realistic chance to craft federal AI rules before the midterm elections. Legal analysts at Fisher Phillips expect no formal advancement before Congress’s August recess; attention now turns to hearings and possible amendments when Congress returns  especially over the preemption’s length (three years versus the ten originally floated) and audit provisions. In the Senate, Marsha Blackburn is pushing her own competing framework with strict child-safety safeguards, while the Trump administration’s December executive order and March AI blueprint keep pressing for federal preemption.

FAQ

1. Is this law yet?

No. It is a discussion draft  released to gather feedback. It still needs formal introduction, passage by both chambers of Congress, and the president’s signature.

2. Does it ban all state AI laws?

No. The proposed preemption applies only to AI model development rules, and only for three years. States keep their authority to regulate how AI is used  employment, privacy, consumer protection and more.

3. Which companies are covered?

“Large frontier developers”  those with more than $500 million in annual gross revenue. Analysts say ordinary businesses and AI users would see little direct impact.

4. Why is opposition so intense?

Critics argue that if the federal standard is weaker than current state laws, it becomes a ceiling rather than a floor  blocking states from responding to new AI harms.

5. What do AI companies stand to gain?

If passed, one national standard replaces fifty state regimes, cutting compliance costs and legal uncertainty  a major factor for investment decisions.

Bottom Line

The Great American AI Act is the first serious attempt to consolidate US AI rulemaking from a state patchwork into one national framework  but the “complete ban on state AI laws” framing is wrong. The preemption runs three years, covers development only, and leaves deployment rules with the states. The real battle is whether the federal standard becomes a floor for safety or a ceiling. Congressional hearings after the recess will decide whether this draft becomes law  or stalls in politics.

Sources / Attribution

1. Politico (June 4, 2026)  draft details, Trahan blowback, midterm context, White House blueprint.
2. Roll Call (June 4, 2026)  3-year preemption, “federalization,” $500M threshold, catastrophic-risk definition, $1M/day penalties, Carson and Steinhauser reactions.
3. Axios (June 4, 2026)  key provisions, CAISI funding, lawmakers’ op-ed quotes.
4. Broadband Breakfast (June 5, 2026)  catastrophic-risk definition detail, whistleblower/education provisions, attorney-general pushback.
5. Daily Signal (June 4, 2026)  Obernolte statement, Trump December executive-order context.
6. Fisher Phillips and legal analyses (June 2026)  deployment rules intact; low odds of movement before August recess.
7. The supplied text’s “complete prohibition” and “Texas” claims could not be matched to primary sources  flagged in Table 3.

All information in this report is source-attributed; as the bill remains at discussion stage, final legislative text may differ. This article will be updated as the bill progresses.