US House Lawmakers Unveil Bipartisan AI Draft: A Three-Year Freeze on State Laws
On June 4, 2026, bipartisan House members released a discussion draft to build a federal AI framework, freeze state laws on AI model development for three years, and require large developers to publish catastrophic-risk frameworks.
Per Reuters and Politico, on June 4, 2026, bipartisan House members Jay Obernolte (R) and Lori Trahan (D) released a discussion draft of AI legislation that would build a federal AI governance framework and lay the groundwork for a federal AI standards center.
The draft's most contentious provision is state-law preemption: it would freeze state legislation targeting AI model development for three years to end the current patchwork of state rules — but explicitly would not apply to local laws governing the 'use and deployment' of AI models. In exchange, the draft would require large AI developers to publish their catastrophic-risk management frameworks, swapping federal transparency requirements for mandatory state-level regulation.
A Continuation of the 'Federal-First' Line
The draft clarifies the core divide in US AI regulation: federal uniformity versus a state-by-state patchwork. Since California's SB 1047 was vetoed in 2024 (see our coverage), the vacuum of 'no federal statute, every state on its own' has left industry struggling with fragmented compliance; companies have lobbied for unified federal legislation to avoid stacking a dozen state regimes. This draft answers that call and continues the innovation-first tone set since the White House AI Action Plan (see our US AI Action Plan coverage).
Compared with the EU's risk-tiered, obligation-heavy approach (see our EU AI Act coverage), the US 'federal preemption + self-disclosure' route leans lighter-touch. Critics worry a blanket freeze on state authority would gut existing local protections in areas like deepfakes and minor safety.
Our Take
As a 'discussion draft,' it is a long way from enacted law, but its signal is clear: the center of gravity in US regulation is not 'how strict' but 'who governs' — the federal-versus-state authority fight is the real battleground. For AI firms, a unified federal framework means lower compliance costs, at the price of accepting catastrophic-risk disclosure; for states that legislated on AI safety early, a three-year freeze means existing rules could be overturned. This 'model development vs. use-and-deployment' jurisdictional split will be the crux of the fight ahead.
This article aggregates official announcements and public reporting; original sources are linked below.
Source:路透社