The Great American Artificial Intelligence Act (GAAIA) is a bipartisan House discussion draft unveiled on June 4, 2026 by Rep. Jay Obernolte (R-Calif.) and Rep. Lori Trahan (D-Mass.). It couples a federal frontier-model transparency mandate with a time-limited federal preemption of state AI-development law, while expressly preserving state authority over the deployment and use of AI.
Status and timeline
The draft was unveiled on June 4, 2026 by Obernolte and Trahan, with Erin Houchin (R-Ind.) and Scott Peters (D-Calif.) as original co-sponsors. It runs 269 pages. On June 4 the Information Technology Industry Council (ITI) endorsed the draft, highlighting its state-preemption provisions. House Democrats and reportedly House GOP leadership both signaled opposition, leaving the draft drawing objections from both flanks. (Source: politico.com; insideaipolicy.com; punchbowl.news; transformernews.ai)
The text is a discussion draft, not introduced legislation, and terms may shift before any markup.
By July 24, 2026, Obernolte and Trahan had formally introduced a successor bill, the Frontier Act, a 74-page frontier-AI framework (Source: insideaipolicy.com; transformernews.ai).
Introduced bill: the Frontier Act (July 2026)
As introduced, the Frontier Act tiers developers by size; requires transparency reports and critical-incident reporting from all tiers; adds annual independent audits and quarterly catastrophic-risk assessments of internal deployments for larger tiers; requires the largest developers to retain Commerce-licensed independent verification organizations; creates a Commerce under-secretary for AI security; permits the Commerce Secretary alone to suspend development or deployment posing imminent catastrophic risk; and narrows federal preemption to state transparency, third-party-audit, and incident-reporting laws (Source: obernolte.house.gov; insideaipolicy.com). Relative to the June discussion draft, the introduced bill is shorter (74 pages against 269), retains the licensed-IVO third-party-audit model, and replaces the draft's three-year development-law preemption with the narrower preemption limited to state transparency, audit, and incident-reporting laws.
Key provisions
The draft requires top AI developers to disclose the safety and security risks of frontier models to third-party auditors. The disclosure regime is auditor-facing, modeled on California SB 53 and New York's RAISE Act, rather than a public-transparency or pre-clearance regime.
It authorizes the Center for AI Standards and Innovation (CAISI), with a $100 million annual budget, to administer the framework. Frontier developers would be required to retain licensed "independent verification organizations" (IVOs) for regular adequacy audits, operationalizing the third-party-auditor model and echoing Dean Ball–style state-authorized-IVO proposals. Misrepresenting safety and security practices or failing to report critical incidents would carry civil penalties of $1 million.
Neither the draft nor the introduced bill specifies what access an adequacy audit confers, and the same holds for its critical-incident duty (see AI Incident Reporting). METR's July 2026 proposal for investigating misalignment incidents sets out what an outside investigator would need to establish causes rather than merely confirm that an audit occurred — the ability to run the models involved, full transcripts or reproducible environments, employee interviews across security, training and internal-investigation staff, and prompted classifiers over the training data — and states that transcript-and-interview access alone "would provide more limited assurance and be insufficient to answer many important questions about the root causes" (How independent researchers could investigate AI propensities after misalignment incidents (METR, July 2026)).
Preemption and the development-vs-deployment line
The draft would bar states for three years from passing laws affecting how leading companies develop AI, a narrower, time-boxed preemption than a blanket moratorium. The preemption section contains an extensive rule of construction exempting AI deployers: states may continue regulating the deployment and use of AI systems while the federal government exclusively oversees development. This sets the draft's federal/state line at the development-vs-deployment boundary. (Source: insideaipolicy.com; draft text: trahan.house.gov)
The structure splits a federal floor on development (transparency to auditors plus a development-law preemption window) from state discretion over deployment and use. That framing distinguishes the draft from several adjacent efforts: the blanket state-AI-law moratorium approaches, including the 2025 budget-reconciliation moratorium that failed and EO 14365's preemption push; the deployment-focused state regimes it would partly displace, such as the Colorado AI Act, Texas TRAIGA, and California SB 53; and the narrower June 2 EO, which concerns government visibility into AI security rather than legislative preemption.
Reactions
Transformer called the draft "the best AI bill yet" while noting it "may not get far." Both House Democrats, citing the preemption of state development law, and reportedly House GOP leadership have signaled opposition, leaving the draft without a clear coalition despite ITI's industry backing. (Source: transformernews.ai)
On June 5, 2026, Trahan published an op-ed defending the framework, as Americans for Responsible Innovation (ARI) ran district ads characterizing it as a stalking horse for state preemption. Separately, David Sacks called the June 2 EO's shortened model-review period a "game changer," linking the legislative-preemption track to the administration's executive-action track. (Source: insideaipolicy.com; insideaipolicy.com) Pro-state-regulation groups attacked the preemption window from one side while parts of the GOP remained wary of any federal mandate from the other.
Relationships
- instance-of: Frontier AI Governance — the federal preemption bargain rung of the instrument spectrum
- related: AI Industry Lobbying — The 2025-2026 Political Offensive — the industry preemption push the draft partly responds to
- related: EO — Trump Federal Preemption of State AI Laws (Dec 11, 2025) — the executive-branch preemption track
- related: Lieu-Obernolte AI Bill (April 2026) — Obernolte's earlier (Lieu–Obernolte) AI bill
- related: EO — Promoting Advanced AI Innovation and Security (Trump, signed June 2, 2026) — the contemporaneous federal AI-security order
- related: NIST CAISI (Center for AI Standards and Innovation) — the agency the draft would fund and empower; AI Federalism — the development-vs-deployment preemption line