Rep. Jay Obernolte (House discussion draft) · 2026
Great American Artificial Intelligence Act (GAAIA)
GAAIA
A House discussion draft from Rep. Jay Obernolte organized into four titles: frontier AI governance, workforce, cybersecurity, and research and international cooperation. It federalizes the SB 53 / RAISE-style transparency model — large frontier developers (>$500M revenue) must write and publicly post a frontier AI framework, file pre-deployment reports, and report critical safety incidents to CAISI within 15 days (or 24 hours if there is an imminent risk of death or serious injury), backed by fines of up to $1M per day and federal and state AG injunctions. CAISI is established in statute within the Department of Commerce to set voluntary security standards and to license Independent Verification Organizations (IVOs) that audit developer frameworks. Its preemption clause is deliberately narrow: it bars only state laws specifically targeting AI model development, expressly preserves laws of general applicability, common law remedies, and regulation of AI use and deployment, and sunsets three years after enactment. Title II builds the most elaborate AI workforce-data apparatus of any proposal — WARN Act layoff disclosures, AI-sensitive occupation forecasts, an AI Workforce Research Hub, and a study of a Rapid AI Adjustment Assistance Program.
Key Provisions
CAISI established in statute within Commerce, setting voluntary AI security standards and licensing Independent Verification Organizations (IVOs) to audit frontier developers
Large frontier developers (>$500M revenue) must publish a frontier AI framework, file pre-deployment reports, and report critical safety incidents (15 days, or 24 hours if imminent), with fines up to $1M/day
Narrow, three-year-sunsetting preemption of state laws targeting AI model development only — preserving common law and regulation of AI use and deployment
WARN Act amended to require disclosure when AI is a 'substantial factor' in a qualifying mass layoff, including the share of job losses attributable to AI
AI Workforce Research Hub, prediction-interval forecasts for AI-sensitive occupations, and a study of a Rapid AI Adjustment Assistance Program modeled on trade adjustment assistance
Whistleblower anti-retaliation protections, increased AI fraud penalties, statutory NAIRR, and international standards coalitions that exclude China absent WTO certification
Regulatory Philosophy
Structured but light-touch federalism that codifies the emerging transparency consensus rather than imposing pre-deployment gates. Obernolte's draft accepts a real federal disclosure-and-audit regime for the largest frontier developers while keeping CAISI's technical standards voluntary, and pairs it with an unusually deep investment in measuring AI's labor-market impact. Its preemption is narrow and time-limited — confined to model development and sunsetting in three years — a marked departure from the broad, permanent preemption sought by the White House and the Blackburn bill.
In contrast
GAAIA vs. the OpenAI Blueprint
GAAIA and OpenAI's Blueprint share an architecture — a statutory CAISI, the SB 53 / RAISE transparency model, and third-party verification (GAAIA's licensed IVOs mirror the Blueprint's certified assessors). They diverge on teeth and scope. GAAIA pairs a soft trigger (disclosure, with no government evaluation gate) with hard enforcement — $1M/day fines and federal and state AG injunctions — and ranges across workforce data, fraud, and education. The Blueprint inverts this: a mandatory CAISI pre-release evaluation (a harder trigger) that is purely advisory, since CAISI can recommend but never block, paired with a single-minded focus on catastrophic CBRN, cyber, and recursive-self-improvement risk that brackets workers entirely. And where GAAIA is operative legislative text, the Blueprint is industry advocacy.
+Federalizes the SB 53 / RAISE transparency model with real teeth — $1M/day fines, federal and state AG injunctions, and licensed third-party IVO audits of developer frameworks
+Preemption is narrow and self-sunsetting (three years, model development only), preserving common law remedies and states' authority over AI use and deployment — a far lighter touch than the White House or Blackburn
+Title II is the most serious workforce effort in the landscape: WARN Act AI-layoff disclosures, AI-sensitive occupation forecasts, and a research hub that would finally generate systematic evidence on displacement
+Strong whistleblower protections — reinstatement, double back pay with interest, and attorney's fees — give insiders a safe channel to report frontier-AI violations
+Pragmatic, bipartisan-flavored construction that codifies CAISI and NAIRR and funds open-source security and testbeds, giving it a more plausible path to passage than messaging-oriented bills
Weaknesses
From the perspective of political opposition
−Transparency and framework audits are not pre-deployment gates — like SB 53 and RAISE, it lets companies ship frontier models first and disclose risks after, with no authority to stop a dangerous release
−CAISI's core standards remain voluntary, and the binding obligations reach only a handful of >$500M developers, leaving the vast majority of AI systems untouched
−The workforce title is overwhelmingly studies, forecasts, reports, and pilots — it measures displacement meticulously but creates no benefits, no redistribution, and only a study of an adjustment program rather than the program itself
−It is silent on the harms states have actually legislated against — child safety, algorithmic discrimination, deepfakes, and copyright — even while preempting a slice of state authority over development
−IVO immunity from loss claims, combined with broad trade-secret and security redactions, could hollow out accountability by letting developers and their paid auditors control what the public and regulators ever see
−The preemption line itself is contestable: it bars only state laws 'specifically targeting' model development while preserving deployment and use regulation, but the development/deployment boundary blurs at exactly the points that matter — pre-deployment safety testing and substantial fine-tuning — inviting years of characterization litigation over which state laws actually survive