The First Real Bipartisan AI Bill Just Dropped — And It’s Already Controversial

On June 4, 2026, Reps. Jay Obernolte (R-CA) and Lori Trahan (D-MA) released the discussion draft of the Great American Artificial Intelligence Act of 2026 — the first serious bipartisan attempt at comprehensive federal AI regulation in the United States. It would create a federal oversight framework, fund it with $300 million over three years, and — here’s the explosive part — preempt state AI laws for three years.

The bill arrives the same week Trump signed his scaled-back voluntary AI executive order and OpenAI published its own divergent policy paper pushing for civilian oversight. The US AI governance train is finally leaving the station. The question is whether it’s heading toward protection or toward a regulatory floor that doubles as a ceiling.

What’s Actually in the Bill

The Great American AI Act has four pillars:

1. Frontier model governance. The bill codifies the Center for AI Standards and Innovation (CAISI) — the office originally created under Biden as the AI Safety Institute and rebranded by the Commerce Department in 2025. CAISI would develop voluntary standards, study national security risks, and oversee a new licensing regime for independent verification organizations — third-party auditors who would confirm that frontier model developers are complying with transparency requirements.

Frontier model developers would need to:

• Craft and publicly disclose an AI governance framework covering all their models
• Identify risk thresholds and determine whether models pose “catastrophic risk”
• Disclose release dates and any modifications to governance frameworks
• Hire qualified independent verification organizations with no ties to the company being audited
• Make audit reports available to state attorneys general on request

2. Workforce impact tracking. The bill creates an Artificial Intelligence Workforce Research Hub at the Department of Labor, directs the BLS to track AI’s impact on employment, and establishes an expert workshop of economists, AI technical experts, industry participants, and labor organizations. It also funds AI education efforts including eight Centers of AI Excellence through the NSF.

3. Cybersecurity. It reauthorizes and extends the Cybersecurity Information Sharing Act of 2015 through 2035, directs Homeland Security to develop outreach plans for small and rural critical infrastructure operators, and funds open-source software maintainers with controlled access to frontier models for vulnerability detection. (Yes, this is a direct response to the Anthropic vulnerability framework release and the University of Toronto AI worm demonstration.)

4. Research and development. Creates a testbed program coordinating national labs, NIST, the NSF, and private sector entities for security risk and vulnerability assessments.

Funding: $100 million per year for 2027-2029, totalling $300 million. Modest, given the stakes.

The Preemption Problem

Here’s where it gets volatile. The bill would preempt state AI laws for three years — meaning California’s SB 1047-style regulations, Colorado’s AI consumer protection law, and any other state-level AI legislation would be overridden by a federal floor.

Public Citizen called it a “disastrous proposal” that Big Tech is “celebrating.” J.B. Branch, AI governance and technology policy counsel at Public Citizen, said the bill “does nothing to address algorithmic discrimination, consumer fraud, youth mental health harms, AI companions, and deepfakes.”

The Tech Oversight Project was equally blunt. Executive director Sacha Haworth called it an “AI amnesty bill,” arguing it “replaces a state floor with a federal ceiling and trades away existing and future child safety, civil rights, and consumer protection laws for nothing in return.”

Polling from the Tech Oversight Project suggests constituents aren’t on board: 56% of Obernolte’s own constituents oppose weakening California’s AI catastrophic risk laws, and 63% of Trahan’s constituents feel the same about local protections.

But here’s the other side: the current patchwork of state laws is genuinely chaotic. Three states passed AI laws in 48 hours in April. Companies operating nationally face a compliance nightmare. Obernolte argues a federal framework provides clarity without smothering innovation. He’s not entirely wrong — the status quo is a mess.

The question is whether this bill fixes the mess or just replaces it with a lower standard.

How It Connects to the Week’s Other AI Governance Moves

This bill didn’t land in a vacuum. The Trump administration signed a scaled-back AI executive order on June 2 — 30-day voluntary review instead of the 90-day mandatory framework he refused to sign in May. That order created CAISI by executive action, but with no funding. The Great American AI Act would give CAISI congressional backing and $300 million.

Meanwhile, OpenAI published its own policy paper on June 3, pushing for civilian oversight through CAISI and NIST rather than NSA-led national security review. OpenAI supports CAISI — but wants it firmly in civilian hands, not military ones.

And across the Atlantic, the EU just unveiled its €320 billion Tech Sovereignty Package with chip emergency powers and cloud sovereignty tiers. The US-EU regulatory divergence is now stark: Europe is building walls; the US is debating whether to build a floor.

💰 Industry Impact

Who benefits: Large frontier model developers (OpenAI, Anthropic, Google, Meta) get a single federal framework instead of 50 state ones. Independent verification organizations — a new category created by this bill — become a whole new compliance industry. Cybersecurity firms get expanded information-sharing mandates and federal funding.

What’s at stake: The AI compliance market is projected to reach $16 billion by 2028. If this bill passes, the independent verification organization licensing regime alone creates a new market segment. Companies like NIST-accredited auditors, existing cybersecurity firms, and startups positioning themselves as AI governance consultancies stand to gain. For NZ tech companies selling into the US market, a single federal framework simplifies compliance significantly — though it may lower the bar below what states like California currently require.

Key risks: The three-year preemption window could leave consumers without state-level protections that are already in force. The $300M funding over three years is modest — the EU is committing €320 billion. And the “voluntary” transparency framework for frontier models mirrors the voluntary approach in Trump’s EO, raising the question: is this regulation or regulatory theatre?

❓ Frequently Asked Questions

Q: What does this mean for New Zealand? A NZ-specific angle: a US federal framework would simplify compliance for NZ tech companies exporting to the US market. But if the federal standard is lower than California’s, it could also mean weaker consumer protections for anyone using AI tools sold under the federal floor. NZ’s own AI strategy — launched in July 2025 — aligns more with the EU’s precautionary approach, so divergence from the US framework could create trade friction.

Q: Does this bill actually regulate AI? It creates a transparency and verification framework for frontier models — but the key word is “voluntary.” Developers must disclose frameworks and hire auditors, but the standards themselves are voluntary, not mandatory. Critics argue this is disclosure without teeth.

Q: What happens to state AI laws? For three years after enactment, the federal framework would preempt state AI laws. After that, states could theoretically pass their own — but the political reality is that once a federal floor exists, it’s very hard to rebuild state-level protections that were stripped away.

Q: Will this pass? It’s a discussion draft, not a bill yet. Obernolte and Trahan are inviting public feedback. The preemption provision alone will face fierce opposition from state attorneys general, consumer advocates, and the growing coalition of states that have already passed AI legislation. But bipartisan AI bills are rare, and this one has real momentum.

🔍 THE BOTTOM LINE

The Great American AI Act is the first serious bipartisan US AI bill — and it reveals the central tension in AI governance: the industry wants federal preemption because state laws are expensive and complicated; consumer advocates want federal regulation that’s stronger than state laws, not weaker. This bill gives the industry most of what it wants. Whether it gives the public enough is the fight that’s about to begin.

SOURCES

• Nextgov/FCW — Lawmakers propose AI framework that would preempt state laws for 3 years
• Gizmodo — New Bipartisan Legislation Takes a Big Step Forward in Restricting State Regulation of AI
• Punchbowl News — Tech: Obernolte-Trahan AI proposal is out
• Public Citizen — Obernolte-Trahan bill strips states’ authority to protect consumers, workers, and children
• ITI — ITI Reacts to the Great American AI Act