Representatives Jay Obernolte, a California Republican, and Lori Trahan, a Massachusetts Democrat, released a 269-page discussion draft on Thursday that would freeze state laws “specifically regulating the development” of artificial intelligence models for three years. The bill, titled the Great American Artificial Intelligence Act and co-sponsored by Representatives Suhas Subramanyam, Scott Franklin, Scott Peters, and Erin Houchin, arrived two days after President Trump signed an executive order steering frontier oversight toward a voluntary federal pathway, and roughly three weeks after Colorado dismantled the country’s first comprehensive state A.I. statute.

The draft draws a sharp line between development and deployment. According to a summary released by Ms. Trahan’s office, California’s AB 2013, which requires developers to publish summaries of training data, would be preempted, as would the content-watermarking portion of SB 942. Frontier safety regimes taking shape in California, New York, and Illinois would be “federalized.” Developers with prior-year gross revenue above $500 million would’ve to publish frontier A.I. frameworks under the federal scheme that replaces them.

That federal scheme is the one Mr. Trump put in motion on June 2. His order, “Promoting Advanced Artificial Intelligence Innovation and Security,” gives the National Security Agency 60 days to develop a classified process for benchmarking models’ cyber capabilities and designating “covered frontier models.” Companies may voluntarily submit those models to trusted federal partners within a 30-day window before public release. NPR reported that an earlier draft contemplated 90 days; the White House shortened it over concerns about competitive pressure from China.

The opposition was immediate and structural, not partisan. Brad Carson, a former Democratic representative from Oklahoma who now leads Americans for Responsible Innovation, called preemption a “generational mistake.” In his statement: “This bill takes the current floor on state AI legislation and turns it into a federal ceiling, preventing state lawmakers from addressing emerging AI harms in an era of fast-moving technology.” Brendan Steinhauser, chief executive of the Alliance for Secure AI, praised the draft’s bipartisan posture but said it “does not justify preempting states’ ability to pass their own AI safeguards.”

The backdrop matters. Colorado passed SB 189 on May 12, repealing the Colorado AI Act in favor of a disclosure-based framework that takes effect January 1, 2027. The state that wrote the first-in-the-nation rulebook walked it back on its own terms, which makes the federal move read less as rescue than as enclosure.

Vendors are already routing around the uncertainty. OpenAI and Anthropic have the legal staff to absorb regime change; the long tail of smaller deployers, many of them building on platforms like LemonLime, has been quietly making it work across whichever deadlines each jurisdiction decides to keep. Practitioners at Lathrop GPM note that the federal approach rests on the premise that existing federal laws are sufficient, a premise several state attorneys general have signaled they’ll contest.

A federal ceiling lasts only as long as the next Congress agrees that it should.

Sources