On June 4, 2026, Representatives Jay Obernolte (R-CA) and Lori Trahan (D-MA) released a 269-page bipartisan discussion draft of the Great American Artificial Intelligence Act (GAAIA) — described by multiple legal trade publications as the first genuinely comprehensive federal AI governance framework to be proposed in Congress, rather than a narrower bill targeting one specific harm like deepfakes or chatbot safety. As of this writing the bill remains a discussion draft, released alongside a joint op-ed from its two sponsors, and has not yet been formally introduced as legislation — the release was explicitly framed as an invitation for feedback from industry, advocacy groups, and the public before a formal bill is filed.
The draft is organized into four titles: Frontier AI Governance, Workforce, Cybersecurity, and Research, Development, and International Cooperation. Its most significant provisions target what the bill calls “large frontier developers” — companies with more than $500 million in annual revenue that have trained a frontier-scale AI model — with binding federal obligations around safety testing, incident reporting, and governance practices. Smaller companies and most applications built on top of frontier models sit largely outside the bill’s most demanding requirements, a deliberate design choice its sponsors have described as aimed at regulating the handful of companies actually building the largest, most capable models, rather than the much larger population of startups building products on top of them.
The Preemption Fight Is the Real Story
The provision generating the most debate is a three-year preemption of state laws that specifically regulate the development of AI models. Under the draft, states would be blocked from passing or enforcing their own AI-development-specific rules for that window, while state laws of general applicability — existing consumer protection, privacy, and healthcare regulations that happen to also cover AI — would remain fully intact, and the preemption wouldn’t extend to how AI systems are actually used after deployment. Critics, including consumer advocacy groups, argue this still meaningfully strips states of their ability to protect consumers, workers, and children from AI-specific harms during exactly the period AI capabilities are advancing fastest. Supporters counter that the current reality — dozens of states independently writing their own, often conflicting AI development rules — is already creating a genuinely unworkable compliance patchwork for any company operating nationally, and that a temporary federal pause buys time for Congress to legislate a durable national standard instead.
This draft builds directly on the White House’s own March 2026 National Policy Framework for Artificial Intelligence, which explicitly recommended against creating any new federal AI regulator and instead called for AI to be governed through existing agencies with relevant subject-matter expertise, alongside federal preemption of state AI development laws. The House Science, Space, and Technology Committee separately marked up and advanced 10 different AI-related bills in June 2026, all with bipartisan support — evidence that appetite for federal AI legislation is real and growing on both sides of the aisle, even though no comprehensive federal AI law has actually been enacted as of this writing.
The Patchwork the Bill Is Trying to Solve Is Already Real
The “unworkable compliance patchwork” GAAIA’s sponsors point to isn’t hypothetical. As of mid-2026, at least five states — California, Colorado, Texas, New York, and Illinois — have comprehensive, cross-sector AI laws either already in force or taking effect by January 1, 2027, each with its own definitions, risk-assessment requirements, and enforcement mechanisms. Layered on top of those five are more than 40 additional, narrower state laws targeting specific AI harms like deepfakes, AI use in hiring, and chatbot disclosure requirements. State AI legislation has also been accelerating sharply: 2025 was the first year every single US state introduced at least one AI-related bill, and roughly 145 of those bills were signed into law across the country that year alone. That’s the real backdrop GAAIA’s preemption fight is playing out against — a genuinely fast-moving, fragmented state regulatory landscape that shows no sign of slowing down on its own.
What This Means for Philippine Founders
Philippine AI and software startups building for the US market — a common target given the size of the English-language SaaS and outsourcing-adjacent opportunity — currently have to think about compliance with individual state AI laws in whichever states their customers happen to be based, an increasingly real burden as more states pass their own rules. If a bill resembling GAAIA is eventually enacted, that patchwork would temporarily simplify into a single federal standard for at least the largest AI model developers, which is good news for smaller Philippine-built products sitting downstream of those models rather than building frontier AI themselves — they’d likely remain outside the bill’s most demanding obligations either way. The more durable lesson for founders right now is less about this specific draft and more about the underlying trend: US AI regulation is moving from state-by-state experimentation toward a real, if contested, push for a unified federal framework, and founders selling AI-adjacent products into the US should expect that landscape to keep shifting over the next one to two years rather than assuming today’s state-by-state rules are the permanent baseline to build a compliance strategy around.
Share this article