Source: Roll Call

Reps. Jay Obernolte (R‑Calif.) and Lori Trahan (D‑Mass.) released a 269‑page discussion draft titled The Great American Artificial Intelligence Act that would preempt state laws “specifically regulating the development” of AI models for three years while leaving state laws on use or deployment intact. The draft would extend the Cybersecurity Information Sharing Act of 2015 through fiscal 2035, and it names California’s AB 2013 (training‑data summaries) and parts of SB 942 (content watermarking) as examples of provisions the bill would preempt or “federalize.”

Large frontier developers — defined by the draft as firms with more than $500 million in gross revenue the prior year — would need to publish public frontier AI frameworks addressing catastrophic risk, cybersecurity of nonpublic model weights, critical safety incident response, and internal‑use risk. The bill sets a statutory catastrophic‑risk test (foreseeable and material risk of death or injury of more than 50 people or more than $1 billion in property damage), would require Independent Verification Organizations licensed through NIST’s Center for AI Standards and Innovation to perform semiannual audits, and exposes developers to up to $1 million per day for safety violations or certain audit failures. The draft would establish CAISI in statute, authorize $100 million per year for fiscal 2027–2029 plus fees from IVO licensing, allow CAISI to hire technical experts at higher pay, and terminate CAISI authorities three years after enactment. Required reporting would include AI critical safety incidents within 15 days and imminent risks within 24 hours, with state attorneys general able to opt into receiving reports; the bill also adds whistleblower protections, penalties for certain AI‑enabled financial crimes and impersonation of federal officials, and tasks agencies with studies and testbeds on AI and content moderation and workforce impacts. The proposal drew immediate criticism for its preemption clause from groups and figures including Brad Carson of Americans for Responsible Innovation and Brendan Steinhauser of the Alliance for Secure AI, who praised bipartisan elements and focus on catastrophic risk but said preemption would hamper states’ ability to enact stronger safeguards and fails to justify blocking state measures that address child safety and consumer protections. Stakeholders expressed concern about aggressive auditing and data‑sharing requirements that could risk trade secrets, and the draft includes multiple education and workforce provisions such as grants for AI literacy and a Labor Department‑convened AI Workforce Research Hub.